1992 TESE NuclearPhenomenainBrazilianPortuguese

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/305775973
Nuclear Phenomena in Brazilian Portuguese
Thesis · January 1992
CITATIONS READS
5 95
1 author:
Thais Cristófaro Silva

Federal University of Minas Gerais
84 PUBLICATIONS 303 CITATIONS
SEE PROFILE
All content following this page was uploaded by Thais Cristófaro Silva on 02 August 2016.
The user has requested enhancement of the downloaded file.

.: .
NUCLEAR PHENOMENA IN
BRAZILIAN PORTUGUESE
by
THAIS CRISTOFARO ALVES DA SILVA
A THESIS SUBMITTED TO THE DEPARTMENT OF

PHONETICS AND LINGUISTICS
IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR
THE DEGREE OF DOCTOR OF PHILOSOPHY
SCHOO~ OF ORIENTAL AND AFRICAN STUDIES
UNIVERSITY OF LONDON
1992
ABSTRACT
NUCLEAR PHENOMENA IN BRAZILIAN PORTUGUESE
In this thesis we will analyse phonological phenomena which involve nuclear

positions in Biazilian Portuguese. The theory of Charm and Government is the
framework on which our analysis is based.
We will discuss the distri-bution of Brazilian Portuguese vowels with respect
to the primary stressed position which, we propose, is subject to charm
constraints imposed on nuclear segments.
Emphasis is placed on analysing vowel coalescence in Brazilian Portuguese.
Specifically, we will consider phonological processes which involve vowe1-
glide alternations. We will argue that whether a high vowel alternates or
fails to alternate with its corresponding glide depends on the governing
relations that adjacent nuclear positions contract with each other.
Our analysis aims not only to.provide a general account for the occurrence of
high vowels and glides, but also to determine how glides are phonolo~ically
interpreted in Brazilian Portuguese.
Department of Phonetics and Linguistics

School of Oriental and African Studies
University of London
~ Thals Crist6faro Alves da Silva
c/o Departamento de Linguistica

Faculdade de Letras
Universidade Federal de Minas Gerais
Campus da Pampu1ha
Belo Horizonte, 31.270
MG, BRAZIL
or
31 Belsize Ave
London NW3 4BL
TABLE OF CONTENTS
ACKNOWLEDGMENTS ------------------------------------------------------- i
PREFACE --------------------------------------------------------------- ii
CHAPTER 1: THE THEORY OF CHARM AND GQVERNMENT
1.0. Introduction ------------------~---------------------------------- 1

1.1. A Theory of Phonological Strings
1.1.0. Introduction --------------------------------------------- 1
1.1.1. On the Notion of Phonological Government.----------------- 1
1.1.2. Constituent Government ----------------------------------- 3
1.1.3. Interconstituent Government ------------------------------ 5
1.1.4. Sequences of Adjacent Nuclear Positions ------------------ 7
1.2. The Internal Representation of Nuclear Segments ------------------ 9 --_ .. .. __.-
..
----l~veriiIng Properties of Nuclear Segments ----------------------~--

_-~--~
14
1.4. Conclusion ------------------------------------------------------- 16
CHAPTER 2: WORD STRESS IN BRAZILIAN PORTUGUESE
2.0. Introduction ----------------------------------------------------- . 17

2.1. Primary Stress Assignment ---------------------------------------- 17
2.2. Secondary Stress and Word Level Stress --------------------------- 25
2.3. Metrical Constraints Imposed on Constituents --------------------- 27
2.4. Conclusion ------------------------------------------------------- 30
CHAPTER 3: THE VOWEL SYSTEM OF BRAZILIAN PORTUGUESE

3.0. Introduction ------------------------------~~--------------~------31
3.1. Primary Stressed Vowels ------------------------------------------ 31
3.2. Posttonic Vowels ------------------------------------------------- 33
3.3. Pretonic Vowels -------------------------------------------------- 34
3.4. Summary of BP Vowel System ---------.------------------------------ 42
3.5. The Underlying Representation of High Vowels --------------------- 42
3.6. Conclusion ------------------------------------------------------- 44
CHAPTER 4: REVIEW OF THE LITERATURE

4.0. Introduction ------~----------------------------------------------45
4.1. The Phonemic View
4.1.1. Glides·as Underlying Segments ---------------------------- 45
4.1.2. Glides Derived from High Vowels -------------------------- 46
4.2. The Linear Generative View
4.2.0. Introduction --------------------------------------------- 47
4.2.1. Glides as Underlying Segments ---------------------------- 48
4.2.2. Glides Derived from High Vowels -------------------------- 49
4.~ The Non-linear View
4.3.0. Introduction --------------------------------------------- 51
4.3.1. Bisol's Analysis ----------------------------------------- 51
4.4. Conclusion ------------------------------------------------------- 55
CHAPTER 5: ON GOVERNING RELATIONS BETWEEN NUCLEAR POSITIONS
5.0. The
Introduction
Obligatory---------,--------------------------------------------
Contour Principle _ 56
5.l. 56
5.2. Previous Proposals on Governing Relations
Between Adjacent Nuclear Positions
5.2. O. Introduction -------------------------------------'-------- 58
5.2.1. Charette's Proposal -------------------------------------- 59
5.2.2. Yoshida's Proposal --------------------------------------- 60
5.2.3. Conclusion ----------------------------------------------- 62
5.3. Internuclear Governing Relations --------------------------------- 63
5.4. Conclusion ------------------------------------------------------- 66
6.0. Introduction ----------------------------------------------------- 67

6.1. The Phonological Representation of Prevocalic
Glides and Prevocalic·High Vowels ------------------~------------ 67
6.2. Pretonic Prevocalic Glides and Pretonic
Prevoca1ic High Vowels --------------------------~--------------- 69
- 6.3.' The Syllabification of the Glide [w] ----------------------------- 79
6.4. Posttonic Prevocalic Glides and Posttonic
Prevocalic High Vowels ------------------------------------------ 85
6.5. Prevocalic Primary Stressed High Vowels -------------------------- 90
6.6. Conclusion -----------~-------------------------------------------95
CHAPTER 7: POSTVOCALIC GLIDES AND POSTVOCALIC HIGH VOWELS
7.0. Introduction ------~----------------------------------------------,98

7.1. The Representation of Vowel-Glide Sequences ---------------------- 98
7.2. The Phonological Interpretation of Vowel-Glide
SeqUences in BP ------------------------------------------------- 99
7.3. Postvocalic Glides and Postvocalic High Vowels ------------------- 105
7.4. Conclusion ------------------------------------------------------- 108
CHAPTER 8: INTERVOCALIC GLIDES AND SEQUENCES OF HIGH VOWELS
8.0. Introduction ----------------------------------------------------- 109

8.1. Intervocalic Glides ---------------------------------------------- 109
8.2. Sequences of High Vowels
8.2.0. Introduction ------~-----------~~-------------------------115
8.2.1. Pretonic Sequences of High Vowels ------------------------ 116
8.2.2. Posttonic Sequences of High Vowels ----------------------- 119
8.2.3. Primary Stressed High Vowels ----------------------------- 123
8.3. Conclusion ------------------------------------------------------- 128
CONCLUSION 129
REFERENCES 131
o resto e mar ...
(TOM JOBIN)
To Johnzinho
ACKNOWLEDGMENTS
The completion of this thesis would not have been possible without help from:
My supervisor, Professor Jonathan Derek Kaye.
My colleagues from the School of Oriental and African Studies and University
College, London, especially Dr. Monik Charette.
Dr. John Harris and Dr. Harry van der Hulst for their comments and suggestions
on various sections of this thesis.
My friends~ espe9ially, Ariadne; Baofa; Barbara; Bernadete; Carlos and Nice;
Cecilia; Ceri; Chico and Cris; Chris and Vania; Dani; Dana; Dona Isa and
family; Evelina; Gracinha and Chris; Isa Terezinha; Krenak people; Lena;
Lucinha; Luiz Carlos; Lysle; Marcio; Marco Antonio; Moyra; Peter; Polly and
lara; Sanzio, Ivana and Natalia; Vic, Help, Carol and Barbara; Zina; Willer
and, of course, Johnzinho.
~
Fabio (he knows why).
The CNPq - Conselho Nacional de Desenvolvimento Cientifico e Tecno16gico - for
grant numbers 20.24.24-86 and 40.24.69.91-7.
i
PREFACE
PREFACE
In this thesis we will analyse phonological phenomena which involve nuclear
positions in Brazilian Portuguese. Specifically, we will consider vowel
coalescence with emphasis on the process of vowel-glide alternations.
The data we present comes from the Brazilian Portuguese dialect spoken in Belo
Horizonte, in the state of Minas Gerais. We will refer to this dialect as BP.
References to other dialects are made whenever they are useful for supporting
e:m ar~ent or when they throw some light on the discussion of a given
lssue.
The analysis we provide is based on nouns and adjectives, whose morphological
structure consists of a noun stern, which may optionally be ~ollowed by a
derivative suffix, in. turn followed by the morpheme of gender.
The framework we have adopted is the theory of Charm and Government which has
been developed by KAYE, LOWENSTAMM and VERGNAUD (1985, 1990).
In Chapter 1 we present the basic assumptions of the theory of Charm and
Government, providing the tools necessary for the analysis we propose later .
. In Chapter 2 we provide a general account of how word stress is assigned in
BP. We will argue that primary stress is lexically determined. We will also
consider how secondary stress and word level stress are assigned and we will
state some constraints which are impos~d on posttonic constituents in BP.
In Chapter 3 we discuss the distribution of BP vowels with respect to the
primary stressed position. We will see that some vowels may only occur in
posttonic posi tion, whereas others may occur ei ther pretonically or in primary
stressed position. We will propose that the distribution of vowels in BP with
respect to the primary stressed posi tion is accounted for by charm constraints
imposed on nuclear segments.
In Chapter 4 we present a review of the literature on high vowels and glides
in BP. We will see that there is controversy as to whether high vowels and
glides are both underlying segments or whether the latter are derived. We will
consider both views in different frameworks showing that further research is
still needed to account for the distribution of high vowels and glides and the
phonological representation of glides in BP.
1 The translation into English of the data originally collected in

Portuguese was based on the examples provided by "Novo Dicionario Barsa das
Linguas Inglesa e Po~uguesa", HOUAISS, A. & AVERY, C.B (ed.). APPLETON-
CENTURY-CROFTS, New York, Division of Meredith Publishing Company, 1964.
2 The processes we consider in this thesis also apply to verbal forms.
However, considering that the presentation of data from verbs would
necessarily lead us to the discussion of verbal morphology - which is more
complex than the morphological structure of nouns and adjectives - we
restricted our data to nouns and adjectives.
11
.: .
PREFACE
In Chapter 5 we consider governing relations involving a sequence of strictly

adjacent nuclear positions. The proposal presented in this Chapter will be
applied to our analysis of vowel-glide alternations in BP which is presented
in the remaining Chapters of this thesis. .
In Chapter 6 we investigate forms in which prevocalic high vowels, i.e.
[iV,uV], and prevocalic glides, _i.e. [jV,wV], occur. We aim to define the
conditions under which a high vowel alternates or fails to alternate with its
corresponding glide. We will also determine how prevocalic palatal glides are
syllabified in BP. We will see that unlike prevocalic palatal glides, forms
-whtch presentprevoca-lic-back glides, e.-g. -rwa-] ~ may- d-Hf-er "with respe-e-t "to
their syllabification.
In Chapter 7 we consider forms which present postvocalic high vowels, i.e.
[Vi,Vu], and forms which present postvocalic glides, i.e [Vj,Vw]. We will
determine the phonological representation of postvocalic glides in BP and also
when a postvocalic high vowel or a postvocalic glide is phonetically
manifested in BP.
Finally, in Chapter 8 we analyse forms presenting intervocalic pal~tal glides
and forms presenting sequences of high vowels. We will determine how
intervocalic glides are phonologically interpreted in BP and which of the high
vowels in a sequence of high vowels may be phonetically manifef?ted as a glide.
iii
CHAPTER 1: THE THEORY OF CHAR! AND GQVERHKEIT
CHAPTER 1
THE THEORY OF CHARM AND GOVERNMENT
1.0. Introduction
In this Chapter we will present the theory of Charm and Government with the
aim of providing the tools necessary for the analysis we propose later. We
will first discuss how the theory addresses the organization of phonological
strings. Then we will consider the internal representation of nuclear segments
in ...Brazilian. .f..Qrt..ugu.e..s.e.. .we .w.ill....al.so.....a.d!1.J'.e..s.s...t.he. .gQy~.~nIDg ..~~QP~.~.t.!~§_P~ .
nuclear segments, d'efining the potential governors and the potential
governees.
1.1. A Theory of Phonological Strings

'--'1.1.0; Introduction
One of the tasks of Government PhOnplogy (henceforth GP) is to explain how
phonological strings are organized. More precisely GP aims to explain how
segments and skeletal positions are associated to phonological constituents.
The theory proposes that skeletal positions along with their segments are
associated to constituents in terms of the relations (called' governing
relations) they contract with each other. The establishment and manifestation
of governing relations are derived from principles of Universal Grammar which,
along with parameters, define phonological systems. In the following section
we will Present the GP proposal concerning the organization of phonological
strings.
1.1.1. On the Notion of Phonological Government
Government is defined as a binary, asymmetric relationship between adjacent
skeletal positions. The establishment of a governing relation between adjacent
positions defines a governing domain where one of the positions is the head -
or the governor - and the other position is the complement - or the governee.
For a governing relation to be established two conditions must be satisfied:
formal and substantive. The ~ormal conditions are stated in (1).
(1) Formal Conditions
a. Strict Locality
b. Strict Directionality
The strict locality condition requires the governor to be adjacent to the
governee at the z~ level of projection. That is, the level of projection at
which all positions are present. The strict directionality condition defines
1 For other lines of research in GP and a full presentation of the theory

as well as theoretical evidence for its basic assumptions see KAYE, J,
LOWENSTAMM, J. &J.-R. VERGNAUD (henceforth KLV) (1985, 1990), CHARETTE (1988,
1991) and KAYE, J. (ed.) (l990b). .. -
1
CHAPTER 1: THE THEORY OF CHARM AND GOVERNMENT
the nature of headship within a governing domain. 2

The substantive condition defines the governing properties of segments, i.e.
whether a given segment is a governor or a governee. The theory assumes that
all phonological segments are formed by a set of primitive, autonomous,
independently pronounceable units which are called elements. These elements
may occupy a skeletal position alone or they may combine, forming a compound
segment. The combinatoriar possibilities of elements are defined in terms of
a property called charm. Besides defining the combinatorial possibilities
of elements, charm characterizes the cardinal nature of segments. Very roughly
speaking, charm characterizes a segment either as a vowel or as a consonant.
Segments _are_ .eLther----charmed-_or __charmless...Charmed._sagments. _.B.re__.e itner
positively or negatively charmed. Positively charmed segments have the
property of "vowe1iness" and they occur in nuclear head positions. Negatively
charmed segments have the property of "consonantiness" and they occur in non-
nuclear head posi tions, i. e. onsets. Charmless segments do not have either the
property of "voweliness" or "consonantiness". Aneutrally charmed or charmless
segment may occur either in a nuclear position or in a non-nuclear position.
The property necessary for a skeletal position to govern or to be governed by
an adjacent skeletal position is given by the charm value of the segments
which are linked to the skeletal pOfi tions. Charmed segments are governors and
charmless segments are governees. According to the theory, phonological or
lexical representations are organized as follows: at the segmental level there
is a linear sequence of segments; at the skeletal level there is a linear
sequence of skeletal posi tions which are associated to segments; at the
consti tuent level there are onset-rime sequences where the rimes are the
immediate projection of ruclear heads. The onset-rime sequence represents what
is called the syllable.
2 Directionality of goverrunent in phonological constituents differs from

syntactic ones in the sense that in the latter directionality is subject to
parametric variation whereas in the former it is universally defined. That is
why di rectionali ty of goverrunent in phonological constituents is called
"strict". This issue will be addressed in detail in the following pages.
3 Later in this Chapter we will discuss the theory of Charm when we
present the internal representation of BP vowels.
4 Later research has shown that the governing property of a segment
involves not only its charm value but also its complexity (cf. KLV (1990».
The notion of complexity and its role in the governing properties of segments ~
is discussed in section 1.3.
5 In GP the nucleus, onset and rime are the only constituents present in
phonological representations. Evidence for denying the so-called Coda
constituent is given in KAYE (l989c). As for the arguments on denying the
syllable as a phonological constituent see KAYE (198ge) and KLv (1990).
2
Nuclear posiVons and their segments are lexically associated to nuclear

consti tuents. As for the non-nuclear positions, they are phonologically
interpreted according to the governing relations they contract wi th each
other. There are two types of governing relations: constituent and
interconstituent. Let us first consider constituent government.
1.1.2. Constituent Government
Consider the Pho~ologicai or lexical representation of [vidrli] 'glass I
illustrated below.
(2) 0 R 0 R
I I
I I
N N
I 1
I I
X X X X X
I I I I I
v- u20
I I I
i + d- r'0
(2) illustrates a sequence of skeletal positions which are associated to

segments. Recall that nuclear posi tions and thei r segments are led cally
associated to nuclear const1tuents. Let us then consider how the non-nuclear
positions are syllabified. Let us first consider the syllabification of the
initial non-nuclear position in (2) - which is filled with the segment v-. The
only avai lable consti t'O.ent to which it may be associated is the initial onset.
Thus, there is no ambiguity r1garding its syllabification, 1.e. it must be
attached to the initial onset. .
Let us now consider the phonological interpretation of the remaining two non-
nuclear positions in (2). In the sequence of non-nuclear positions illustrated
in (2) the far left position is filled with a charmed segment, i.e. the
negatively charmed segment d-. The non-nuclear position which immediately
follows d- is filled with a charmless segment, i.e. the neutrally charmed
6 This property follows from the "Licensing Principle~ which establishes

that all positions in a domain must be licensed, except the head of this
domain (cf Kaye (198ge». Nuclear positions are the head of the onset-nucleus
sequence so that they are themselves licensed.
7 The symbols g and 1 correspond to lax .high vowels and their
counterparts! and y correspond to tense high vowels. The superscripts on the
segments illustrated in (2) correspond to the charm value of those segments.
The role of charm in phonological representations is addressed in the
following pages.
8 In the cases where a sequence of non-nuclear positions occurs,
syllabification may be apparently ambiguous. This is because in a sequence of
non-nuclear positions both positions may be associated to the same
constituent, i.e. a branching onset, or to different constituents, i.e. a
rimal position and the following onset. As we will see shortly, in GP there
is no ambiguity regarding the syllabification of a sequence of non-nuclear
positions.
3
CHAPTER 1: THE THEORY OF CHARK AND GOVERNMENT
segment rOo We have seen that charmed segments are governors and charmless
segments are governees. Therefore, the negatively charmed segment d- must
govern the charmless segment rOo Given that the non-nuclear position filled
with d- has to govern the non-nuclear position filled with r O, it must be
associated to a constituent head. This is because only a head position. may
govern a complement. In order for the skeletal position filled with d- to be
associated to a constituent head it is projected over the onset constituent.
Given that d- is associated to a constituent head, it can govern the following
non-nuclear position which is filled with rOo We derive a structure where two
positions interacting in a governing relation are associated to the same
constituent.
(3) 0 R 0 R
I
,l\ \
I
1 I
N I N
I I
I \ I
X X ~ X X
I I I I
v-
I I I I
i+ d- r O aO
I
I i
A governing relation involving two adjacent positions which are associated to
the same constituent is defined as:
(4) Constituent Government
a. Strictly local
b. Strictly directional - head initial
governee at the zero level of projection. The strict directionality condition
requires the head to be initial in all phonological systems. Constituent
government operates from left-to-right. It follows from the conditions in (4)
that constituents are maximally binary, i.e. franching constituents are at
most associated to two skeletal positions. In (5) we illustrate the
branching constituents (the head is underlined).
(5 ) a. 0 b. N c. 1lI
I I
1\ I \ I \
I I I
I \ I \ I \
I
I \ I
I \ N .\
! x ! x I
I. \
I I I I
X :It
a-
I I
a~/o
I
~o ~o ..... I
I
a+/ o '0
(3
(Sa) represents a branching onset governing domain where the head presents a
negatively charmed segment and the comp4leffient must present a neutrally charmed
segment. Branching onsets present a stop or non-sibilant fricative in the head
position and a liquid in the complement position. BP is a language which
9 This claims follows from the "Binary Theorem" which is demonstrated in

KLV (1990).
4
CHAPTER 1: fBE THEORY OF CHARH AND GOVERNMBNT
presents branching onsets, ego [klaru] 'clear' or [livru] 'book' .10

(5b) illustrates a branching nucleus governing do~rin. It represents the
structure of both long vowels and heavy diphthongs. Long vowels represent
the case in which the same segmental material occupies both the governing and
the governed posi tions. Heavy diphthongs represent the case where the
governing position is filled with either a positively charmed segment or with
a complex charmles~2segment and the governed position is filled with a simplex
charmless segment. .
(5c) represents a branching rime governing domain where the nuclear head
governs its following rimal complement. The theory proposes that nuclear head
positions have the property of governing non-nuclear positions regardless or------
their charm value (cf. KAYE (198ge». Therefore, nuclear heads always govern
rimal complements. The segment occurring as the nuclear head may be either
posi ti vely charmed or charmless and the segment occurring in the rimal
complement will be neutrally charmed. Branching rimes occur in BP in forms
like [pastu] 'field' and [p6htu] 'harbour'.
The representations in (5) illustrate governing domains 1n which·two skeletal
positions are associated to the same constituent. In the following section we
wi 11 consider governing domains involving skeletal positions which are
associated to different constituents.
1.1.3. Interconstituent Government
Consider the lexical representation of [pastu] 'field' illustrated below:
(6) 0 R 0 R
I I
I I
N N
I I
I I
X X X X X
I I I I I
I
P
- I I
a+ sO t- uO
I I
From the lexical representation in (6) we must show to which constituents the
non-nuclear posi tions are associated. The far left non-nuclear posi hon filled
with the segment p. is associated to the initial available onset. As for the
two other non-nuclear positions filled with SO and t-, we again must show to
which constituents they are associated. Non-nuclear skeletal positions along
with their segments are associated to constituents according to the governing
10 A phonological process involving branching onsets in BP~s discussed

in Chapter·6.
11 Heavy-diphthongs represent what has been referred to in the literature
as falling diphthongs.
12 The internal representation of nuclear segments in BP is defined later
in this Chapter.
5
CHAPTER 1: THE THEORY OF CHARM lXD GOVRRXMRXT
relations they contract with each other. Therefore, the non-nuclear positions
in (6) will be associated to constituents according to the governing relations
established between them.
In the preceding section we considered the form [vidru] 'glass' showing that
its sequence of adjacent non-nuclear positions - filled with the segments d-
and r O - are associated to the same constituent (cf. (3)). In this case the
first non-nuclear position in the sequence is filled with a negatively charmed
segment, Le. d-, being followed by a non-nuclear position filled with a
neutrally charmed segment, i.e. rOo In (6) we also have a sequence of non-
nuclear positions. However, unlike the case of [vidru] - where the sequence
of non-nuclear positions presents a negatively charmed segment followed by a
---n~tLaH-y-eha-rmed-se-gment--the-rep-resent-ation-i-n-{-6i--presen-t8-a-s-equenee-of--------------
non-nuclear positions where a neutrally charmed segment, i.e. so, is followed
by a negatively charmed segment, i.e. t~.
We have seen that charmed segments are-governors and charmless segments are
governees. Therefore, the non-nuclear position filled with t- in (6) will
govern the position filled with so. In order for the non-nuclear position
filled with t- to govern the non-nuclear position filled with SO it must be
associated to a constituent head. This is because only a head position may
govern a complement. The non-nuclear position filled with t- is then projected
over the onset constituent - which is a non-nuclear head. Given that t- is
associated to a constituent head, it can govern the preceding position filled
with SO which is then associated to the rimal complement (which is a non-head
position). We derive a structure where two positions interacting in a
governing relation are associated to different constituents.
(7) 0 R 0 R
:\ I
I
I
I
N\ I
I N
I I J
I \ I I
X X X ~ X
I I I I I
I
p-
I I I I
a+ SO t- o.0
I
1 I
A governing relation involving two adjacent positions which are associated to

different constituents is defined as:
(8) Interconstituent Government
a. Strictly local
b. Strictly directional - head final
governee at the zero level of projection. The strict directionality condition
requires the head to be final i~all phonological systems. Interconstituent
government operates from right-to-left. In (9) we illustrate an
interconstituent governing domain (the head is underlined).
6
CHAPTER 1: THE THEORY OF CHARM AND GOVERXKRNT
(9) 0 R 0
I
1\ 1
N\ I
t
I I
I \ I
X X X
I
~o
I 1
a-
t t
I
(9) illustrates a non-nuclear interconsti tuent governing domain. According to

the theory, a governing relation is always present between an onset and the
. -.preceding._rimal_.comp.l.ement_tcf.._.KA'lE__Cl9_8.9_cJ.L._T..he....onseLp-osi t ion usua!ly_ .._..
presents a negativ~ly charmed segment and the rimal complement presents a
charmless segment. 1 8P is a language w~!cn presents branching rimes, e.g.
[pastu] 'field' and [tahdI] 'afternoon'. .
In this section we considered interconstituent governing domains, i.e.
governing domains in which the adjacent positions are associated to different
constituents. In this thesis we will be investigating governing relations
which involve sequences of strictly adjacent nuclear positions. The following
section is devoted to the investigation of representations presenting such
sequences.
1.1.4. Sequences of Adjacent Nuclear Positions
In this section we investigate governing relations involving strictly adjacent
nuclear positions. Consider the representation in (10).
(10) 0 R 0 R
I
I ,
I
N N
I
I
X X
,
I I
I
a ~
13 A governing relation between an onset and a rimal complement still

holds when both positions are filled with neutrally charmed segments. In this
case the neutrally charmed segment occurring in the onset position must have
a complexity greater than the neutrally charmed segment occurring in the rimal
complement. For ·details on the charm value and complexity of segments
occurring in non-nuclear head positions see HARRIS (1990) .
..,.
14 In some dialects of 8P, including the 8elo Horizonte one, there is a
phonological process according to which an alveolar stop is phonetically
manifested as an alveo-palatal affricate when it is followed by a front high
vowel, Le. /tia/ --> [eia] 'aunt' and Idia/ --> [d~ia] 'day'. For typing
purposes forms presenting alveolar stops followed by a high vowel will be
phonetically transcribed with an alveolar stop rather than with an alveo-
palatal affricate, i.e. [tahdi] rather than [tahd~i).
7
CHAPTER 1: fHE fHEORY OF CHARM AND GOVERNMENT
In order to consider the representation illustrated in (10) - which involves

a sequence of adjacent nuclear positions - let us examine the phonological
representations of nuclear and non-nuclear positions.
Nuclear heads are associated to skeletal positions at the level of lexical
representation (ci. 1.1.1). A nucleus although it always dominates a skeletal
position mayor may not have segmental content. That is, a nuclear skeletal
point may be linked to a segment (cf. (llal~ or it may have no segmental
content, i.e. it may be "empty" (cf. (lIb».
(11 ) a. 0 R b. 0 R
I I I I
I I I I
-- :-----N--- --- ----I --N----------
I I I I
I I I I
X X X X
I , I I
I I I I
~ a ~ 0
Onsets on the other hand can either be associated to a skeletal position
filled with segmental material (cf. (l2a»), or they can be associated to a
skeletal position which dominates no segmental faterial (cf. (12b», or they
can dominate no skeletal position (cf. (l2c».1
(12) a. 0 R b. 0 R c. 0 R
I I I I I
I I I I I
I N ,I
N N
I I I I I
I I I I I
X X X X X
I I I I I
I I I I t
~ a 0 a 0(
Let us now examine the governing relations which involve a sequence of

adjacent nuclear positions - as illustrated in (10). Notice that the onset
which intervenes between the adjacent nuclear positions in (10) dominates no
skeletal position, so that the two nuclear positions are adjacent and
therefore subject to government. Remember that for a governing relation to
hold between two adjacent positions one of the positions must be filled with
a segment which has the property necessary for governing the segment occurring
in the other position. The property of.a skeletal posl tion to govern or to be
15 The theory claims that "empty" positions are filled with an element
which presents special properties. This element is called the cold element -
which is represented by vo. The properties of the cold element are addressed
later in this Chapter when we present the internal representation of nuclear
segments. Empty positions ate subject to a special type of government called
Proper Government. For the discussion of Proper Government see KAYE (l990a),
CHARETTE (1988, 1991).
16 We refer the reader to CHARETTE (1988) for the discussion on the
different behaviour of onsets which dominate no segmental material (cf. (l2b»
and onsets which dominate no skeletal position (cf. (l2c».
8
governed is given by the charm value of the segments which are linked to the.
skeletal positions. The charm value of a segment is given by its internal
representation. The internal representation of a segment consists of the
elements of which it is formed.
In this thesis we will be concerned with governing relations which may hold
in representations such as ~e one illustrated in (10), i.e. sequences of
adjacent nuclear positions. In order to consider governing relations
invol ving adjacent nuclear positions we have to determine the governing
properties of nuclear segments. That is, we have to define which nuclear
segments are potential governors and which ones are potential governees. In
---------the-to.ll.ow-i.ng-s.ectionsJl_a.wil LdiscJ1SS the internal representation of nuclear
segments defining the class of governors and the class of governees. ---------
1.2.The Internal Representation of Nuclear Segments

In this section we will present the theory of phonfilogical representation of
vowels as proposed in KLV (1985) and KAYE (198ge). According to the theory,
any segment consists either of one element or a coinbination of elements.
Elements are understood as autonomous, independent pronounceable units. It is
assumed that any vowel system is derived from the set of elements given in
(13).
The superscripts assigned to the elements in (13) represent their charm val ue.
Charm is a property WhichigamOng other things, expresses the combinatorial
possibilities of elements.
Elements occurring in nuclear positions are either positively or neutrally
charmed. Positive charm represents the property of "voweliness". Since what
characterizes a vowel is resonance, an element which enhances a particular
resonating cavity is positively charmed. In (l4) we illustrate the positively
charmed elements from (13) and the respective resonating cavity they enhance.
(14) Oral cavity
Pharyngeal cavity
Nasal cavity
17 In Chapter 5 we provide a theoretical discussion on governing

relations between adjacent nuclear positions.
18 For the internal representation of consonants~f. KLV (1990), YOSHIDA
(1991), HARRIS (1990).
19 Charm also determines cardinal properties of segments. Positively
charmed segments are associated to nuclear head positions. Negatively charmed
segments are associated to non-nuclear head positions, i.e. onsets. Neutrally
charmed or charmless segments are associated either to nuclear or non-nuclear
positions.
9
The remaining elements in (13), Le. 1°, tfl and vo, are neutrally charmed, Le.
charmless. A neutrally charmed element has neither the property of
"voweliness" nor "consonantiness". That is, a charmless segment can be
assigned ei~her2~0 a nuclear position or to a non-nuclear position, 1.e. an
onset or a rIme. The theory claims that any vowel system is derived from the
set of elements given in (13). Those elements may occur either alone or in
combination. In order to combine elements the "Charm Principle" must be
satisfied .
. (15) Charm Principle
Elements with like charm cannot combine.
-AccoTdi-ng-t(')~he-eha-rm-Pri-ne·i-pte,·t-he--pos-i-t-i-ve-l-y-eh-a-rmed.-eternen-t-s---Ai,-.... i.,-_·_·- _._. __...
N+ - cannot combine with each other since they have like charm. Notice that
neutrally charmed elements - 1°, tfl, V o - may combine with each other. This is
because the charm property of a neutrally charmed element is the absence of
charm so that the charm principle is not violated. The combination of elements
is given by a phonological expression which reflects a fusion process. In (16)
we illustrate a combination of two elements.
(16) (A+. 1°)
I I
I I
operator head
A phonological expression always involves a binary operation. The full stop
in (16) separates the two terms of the phonological expression. The term on
the ~1ft is called the operator and the -term on the right is called the
head. Each element has a number of pnonetic properties one of which is
marked. TEe marke.d phonetic property of an element is called its salient
property. In (17) we illustrate the elements presented in (13) and their
respective salient properties.
(17) A+ openness (non-high)
1° frontness (non-back)
tfl roundness
...+ ATR-ness
Nt nasality
V
o none
In a phonological expression the element which is the operator is not fully
expressed; only its salient property is. In a fusion process the element which
is the operator transmits its salient property to the element which is the
20 In Chapter 6 we will address in detail the representations in which

the charmless elements 1° and trO occupy a nuclear position or a non-nuclear
posi tion. .,.
21 Henceforth, in any phonological expression we will underline the
element which is the head.
22 In KLV (1985) the salient property of an element is referred to as its
"hot feature".
10
CHAPTBR 1: THE THEORY OF CHARK AND GOVERNMENT
head. The segment which results from a fusion process has all the properties
of its head except for the salient property of its operator. Consider the
phonological expressions in (18).
(18) a. (A+ 1°)0 = [E]
b. (1° A+)+ = [at]
Let us first consider (18a). The operator in (l8a) is the element A+ whose
salient property is openness. Given that in a fusion operation the operator
transmi ts its salient property to the head, the property of openness is
-tr.ansmi-tt.ed...to.the-elementI~ which.J.s_the'Hhead._of_the....phonolD.gical .expr.ess.i.on
in (18a). The result of the fusion process illustrated in (18a) is an 1° which
has the property of openness, i.e. [E]. In (18b) the operator is the element
1° whose salient property is frontness. The property of frontness from the
operator is transmitted to the head of the phonological expression in (18b),
Le. A+. The result of the phonological expression in (18b) is an A+ which has
the property of frontness, i.e. [~).
Notice that outside the parentheses which enclose the phonological expressions
in (18) there is a superscript. This superscript corresponds to the charm
value of the phonological expression, i.e. the charm value of a s~ent. The
charm value of a segment is given by the charm value of its head. In (18a)
the head is the element 1° which is 'neutrally charmed so that the charm value
of the segment resulting from the phonological expression in (18a), i.e. [El,
is neutrally charmed. In like manner the segment which results from the
phonological expression in (lab), i.e. [~), is positively charme~ since the
head of the phonological expression in (18b) is the element A+ which is
positively charmed. Let us now consider phonological expressions which involve
more than two elements. Consider (19).
(19) illust!rtes a phonological expression whose head is another phonological

expression. The result of the phonological expression which is the head in
(19) is the segment [El - whose charm value is neutral (cf. (18a». The
operator of the full expression in (19) is the ATR element which gives its
salient property, i.e. ATR-ness, to the head. The resultant segment from the
phonological expression in (19) is a tense mid vowel, Le. [e]. It is
important to mention that the ATR element plays a special role wtth respect
to charm. It functions as i f its salient property; I.e. ATR-ness, gives
positive charm value to a phonological expression (cf. KLV (1985:312)).
Therefore, any phonological expression presenting the ATR element is
23 The positively charmed element ..+ anar It have the property of

transmitting their charm value when they occur as operators in a phonological
expression. We will address this issue shortly.
24 The theory claims that the head of a phonological expression can be
another phonological expression - as illustrated in (19) - but the operator
must be an element. It follows that a phonological expression such as
*((A+.Io).~+) is not possible.
11
positively charmed. Notice that the segment resulting from the phonological
expression in (19), i.e. [e], is positively charmed.
In (17) we have illustrated the salient property of the elements from which
vowel systems are derived. Notice that there 1s an element, i.e. vo, which
does not present any salient property. This element is called the "cold
element" or the "cold vowel". Given that the cold element does not have any
salient property, it cannot be open - which is the salient property of A+; it
_cannot be front - which is the salient property of 1°; it cannot be round -
which is the salient property of UO; it cannot be ATR, which is the salient
property of .+; and it cannot be nasal - which is the salient property of N+.
Therefore the cold element must be a high, back, unrounded, lax and non-nasal
vowel, Le. [-']. Recall that in a phbnorogTcal expre-ssioIC tne resulting-
segment receives all the properties of its head except for the salient
property of the operator. If the cold element occurs as the operator in a
phonological expression it has no role to play since the cold element does not
have any salient property to transmit in a fusion operation. For example, the
phonological expression (vo .~+)+ is interpreted as the same as the single
element A+, i.e. both expressions correspond to the segment [a]. This is
because in a phonological expression such as (vO.~+)+ the operator, i.e. the
cold element has no salient property to transmit to the head, Le. A+.
Therefore (vO.~+)+ is interpreted as [a]. On the other hand, if the cold
element is the head of a phonological expression, e.g. (A+.vO)o, it has a role
to play. That is, in a phonological expression such as (At.yo)o the salient
property of openness from the operator, i.e. A+, is transmitted to the head,
i.e. the cold element. The segment which results from (A+.yO)o is an open [~]
which corresponds to a schwa, i ,e. [a!.
Notice that the segment [a] is
neutrally charmed since its head, i.e. y, is a neutrally char~ed element.
As we have seen, the cold element only plays a role in a phonological
expression when it occurs as the head. If the cold element occurs as the
operator, itnas no effect over the phonological expression. The cold element
may be understood as the number zero in the representation of numbers in the
decimal numerical system. Notice that "three" can be represented either as 3
or as 03. This is because the number zero has no effect over the
representation of decimal numbers when it occurs to the left of a given
number. In like manner the cold element may be understood as being present as
the operator in a phonological expression. That is, a phonological expression
such as (vo.~+)+ is identical to A+ given that the cold element has no effect
over the phonological expre~sion when it occurs as the operator. Consider the
phonological expressions in (20) which represent the BP oral vowels:
(20) a. A+ or (vo .~+)+ = [a]
b. 10 or (vo.lo)o = [i)
c. u<> or (vO.YO)o = [u]
d. (A+ .10)0 = [E]
4 ' (A+ .:!f)o
e.
(A+ .yo)o
= [::>1
f. = [a]
g. (.i'+ .10 )+ = [i]
h. (~+ .~)+ = [u]
1. (~+ .(A+ .1°)0)+ = [e]
j. (Jr+ .(A+ .~)o)+ = [0]
12
CHAPTER 1: THE THEORY OF CHARM AND GOVERN~EHT
The segments [a,i,u], correspond to the elements A+, 1°, uP, pronounced in
isolation or with the cold element as the operator. The segments [E,~,a,i,u]
result from a phonological expression involving a simplex head and an
operator. The segments [e,o] result from a phonological expression involving
a complex head and the ATR element as the operator. In (20) we presented the -
internal representation of BP oral vowels. Nasal vowels also occur in BP as
illustrated in the forms below.
(21) a. [14J 'wool'
b. [ku~tru] 'coriander'
c. [kofbra] 'Coimbra'
______________________d~_Ln.e.6J-- ~.o~_ _
e. [piQ.] 'black fly'
f. [pi4nu] 'piano'
g. [vi~na]
, 'Vienna'
h. [haIiia] 'queen'
1. [mi5ma]
,
j. [siG.ml]
. 'myoma'
'jealousy'
Nasal vowels in BP are derived from vowel-nasal consonant sequences where the
nasal consonant mayor may not surface. (2la-e) illustrate cases where the
nasal consonant does not surface, e.g. /lan0/ --> [laJ and !kuEntru! -->
[ku~tru]. (21f-j) illustrat~5the cases the nasal consonant surfaces, e.g.
/pianu! --> [pianu] 'piano'. .
According to the theory, the nasal element ~ can only combine With neutrally
charmed segments. This constraint follows from the Charm Principle which
establishes that elements with like charm cannot combine (cf. (15». Given
that the nasal element is positively charmed it cannot combine wi th post ti vely
charmed elements. Considering the set of BP oral vowels illustrated in (20)
one expects only the neutrally charmed segments, l.e. [i,u,a,E,::>J, to co-occur
with the nasal element ~. This is exactly the c~se. In (22) we illustrate the
internal representation of nasal vowels in BP.
(22) a. (N+ . (A+. yO ) °)+ = eli]
b. {N+ . (A+.1°)°)+ = [~]
c. (If. (vo .1°)0)+ = [i]
d. (N+ .(A+ .YO)o)+ = [0]
e. (If. (vo .YO)o)+ = [0.]
Notice that when the low vowel [a] is nasalized, e.g. !kama/ --> [kdma] 'bed',
it is phonetically manifested as a nasal schwa, i.e. [aJ, rather than as a
nasal open vowel, i.e. *[a]. The fact that the vowel [a] is phonetically
manifested as a schwa when it is nasalized follows from the theory of Charm.
As we have seen [a] is a positively~harmed segment whose internal
25 For descriptions and analysis of nasal vowels in BP see CAMARA

(1970), ALMEIDA (1976), CAGLIARI (1977,1982), SHAW (1986) AND SILVA (1989a).
26 For typing purposes the lax vowels [i, u, E,::>] are transcribed as
[i,G.,e,o] when they are nasalized.
13
CHAPTER 1: raE THEORY OF CHARK AND GOVERNMENT
representation is (VO.~f)+. Given that the segment [a] is positively charmed

it cannot combine with the nasal element which is also positively charmed.
This is because, according to the Charm Principle, elements with like charm
cannot combine. Therefore, a phonological expression such as *(N+.(VO.~f)+)f
cannot be phonologically interpreted given that the head and the operator are
both positively charmed. However, if the head and the operator of the
phonological expression (VO.~f)+ - which corresponds to the segment raj - are
flipped around we have a neutrally charmed segment, i.e. (A+ .yO)o, which
corresponds to a schwa. Given that the .schwa is a neutrally charmed segment
it may combine with the positively charmed nasal element. That is, we have a
phonological expression such as (Nf . (Af .yO)O)f which corresponds to a nasal
schwa, i.e. [a]. In (22) we illustrated the internal representation of BP
··--nasa-l---vewe-l-s-.--I-n--Ghapte·l'---3----we--wi-l-l-een-s-i:E1e-I'--seme---phene-l-e!1l-i:ea-l--phenemena- -
inval ving the nasalization of vowels followed by a nasal consonant which
surfaces, e.g. /kama/ --> [kama] 'bed'.
In this thesis we will be concerned with phonological processes involving
governing relations between adjacent nuclear positions. In order to consider
governing relations involving adjacent nuclear positions we have to determine
the governing properties of nuclear segments. That is, we have to define which
nuclear segments are potential governors and which ones are potential
governees. In the following section we will define the class of the governors
and the class of governees. .
1.3. Governing Properties of Nuclear Segments

In this section we will discuss governing properties of nuclear segments in
order to define which of ~hese segments are potential governors and which ones
are potential governees. 2 According to the theory the necessary property for
I
a skeletal position ta govern or to be governed by an adjacent skeletal

position is given by the charm value and the fa0mplexity nature of the segments
which are linked to the skeletal positions. Consider (23).
(23) a. Charmed segments may govern; charmless
segments may be governed.
b. Charmed segments may not be governed.
, c. Charmless segments may govern if they have a complexity
. greater than their governee.
(23) illustrates the principles which define governing properties of segments
(cf. KLV (1990)). According to these principles we are able to define which
~ 27 We will restrict the discussion of governing properties of nuclear

segments to the BP nuclear segments whose internal representations were given
in the preceding section.
28 The complexity of a segment concerns the number of elements of which
it is made (cf. section l. 2.1). The role of complexi ty wi th respect to
governing properties of segments will be discussed shortly.
14
.: .
BP nuclear segments are potential governors and which ones are potential
governees. In (24) we i 11ustrate the BP oral vowels and thei r respective charm
value and complexity nature.
(24) Positively Charmed SeFents
a. [a] A
b.
c.
d.
[e]
[i]
[0]
(..+.1° r
( ..+.(A+ _1°)°)+
(..+•("A. yo )° )+
e. [u] (..+.!f)+
Simplex Neutrally Charmed Segments

1. [i:] 1°
j. [u] UO - ---
(24a-e) illustrate positively charmed segments and (24f-j) illustrate
charmless segments. The charmless segments illustrated in (24f-j) may be
either complex (cf. (24f-h», i.e. they result from a combination of elements
in a fusion 'process; or they may be simplex (ct. (24i-j», i.e. they are
formed by a single element. According to the theory, charmed segments may
govern and charmless segments may be governed (ct. (23a». Therefore, t~l
positively charmed segments illustrated in ·(24a-e) are potential governors
and the charmless segments illustrated in (24f-j) are potential governees.
Recall that charmed segments may not be governed (cf. (23b». Therefore, the
positively charmed segments given in (24a-e) cannot be governed. Charmless
segments on the other hand may either be governed (by a charmed segment) or
they may govern another charmless segment. The possibility of a charmless
segment governing another charmless segment depends on its complexity. The
complexity of a segment concerns the number of elements which are present in
its internal representation. That is, the greater the number of elements in
the internal representation of a segment, the more complex it is. Therefore,
a charmless segment governs another charmless segment if it has a complexity
greater than its governee (cf. (23c». It follows from this that the complex
charmless segments illustrated in (24f-h), i.e. [E,~,a], have the property of
governing the simplex charmless segments illustrated in (24i-j)' Le. [i, u] .
This 1s because complex charmless segments - whose internal representations
present two elements - have a complexi ty greater than simplex charmless
segments - whose internal representations consist of a single element. It
follows from this that neither of the simplex charmless segments [i] and [u] -
which have the same d~ree of complexity - should govern the other.
Nevertheless, this is not the case. Indeed a simplex charmless segment may
govern another simplex charmless segment. Governing relations involving the
simplex charmless segments [i] and [u] are discussed in detail in Chapter 8.
29 Recall that nasal vowels are positively charmed segments (cf. (22»
so that they are also potential governors.
15
i'
In (25) we illustrate the governors and the governees of nuclear segments in

BP.
(25) a. Positively Charmed Governors: [a,e,i,o,u]
Charmless qovernees: [E,~,a,i,uJ
b. Complex Charmless GovernQrs: [e,~,a]
Simplex Charmless Governees: [i,ti]
(25a) shows that positively charmed segments may govern either complex or
simplex charmless segments. (25b) shows that complex charmless segments may
govern simplex charmless segments. In this section we have defined the
----governors an(t-the-goveYm~e.~--of~P-nuctea.r --s-e-grnent-s----so-""'that--we-are-ab-I-e--to- .._- ... -.---.. .
consider the governing relations involving a sequence of adjacent nuclear
posi tions.
1. 4. Conclusion
In this Chapter we have presented the theory of Charm and Government. We have
defined how phonological government operates and we have shown how skeletal
positions are syllabified. We have also defined the internal representation
of BP nuclear segments and we have determined the class of governors and the
class of governees.
Throughout this thesis we will be investigating phonological phenomena in BP
which involve nuclear head positions. Nuclear heads are potentially subject
to bear stress. In BP phonology the role of word stress is accorded a
particularly significant place. Therefore, in the following Chapter we will
discuss how word stress is assigned in BP.
16
CHAPTER 2: WORD STRESS IN BP
CHAPTER 2
WORD STRESS IN BP
2.0. Introduction
The role of word stress in BP phonology is very important since various
processes are sensitive to stress placement. Therefore; in order to consider
the processes we will address later in this thesis, we have to·show how stress
is assigned. In this Chap er we aim to provide a general account of how word
stress is assigned in BP. 1 We will argue that primary stress is lexically
determined. We wi 1L.al.s.o-C.ans~.s..ec.ondarLSires.s---and......w.QrQ. level stress
are assigned. Finally, we will state some constraints imposed on posttonic
constituents in BP.
2.1. Primary Stress Assignment

In this section.we will determine how primary stress is assigned in BP.
Primary stress may be either final, penUltimate or antepenultimate. In (1) we
illustfate these patterns. The acute accent marks the primary stressed
vowel.
1 There are few works on BP which deal explicitly with stress assignment.
Among these are: MAJOR (1981, 1985) and SEGUNDO (1990, in preparation). In
this Chapter we refrai.a from discussing previous analyses of stress assignment
since we aim to provide a general account of stress placement rather than to
provide a detailed theoretical discussion on the topic. Nevertheless, we will
assume some of SEGUNDO's (in preparation) proposals throughout this thesis.
There· are few forms where primary stress may fallon the fourth-to-
last nuclear position, e.g. [hitimika] 'rhythmic'. We will consider these
forms later in this Chapter.
17
.: .

""
iii. Antepenultimate stress

a. [mAkina] 'machine'
b. .£Elisi] 'helix I
c. [ezitG.] 'success'
d. [sHaba] 'syllable'
e. [ka~tikG.] 'chaotic'
f. [kr6nikG.] 'chronic'
g. [uviila] 'uvula'
Forms in (l) show that final, penultimate and antepenul timate stressed nuclear
- ----Pos-i-t-i-Gn.s--may.-be··-··-fi-l-l-ed-w-i-t-h----any.-nu-e-l-ea-r--segmen-t--·-f-rem--t-he--se-t-:-[-a,-e,E--,-- - ..-.-.----
i,o,~,u].3 In order to address how primary stress is assigned in BP we have
to consider the morphological structure of nouns. This is because nuclear
positions which are likely to bear primary stress must belong to a noun stem
or to a suffix other than the suffix for gender. Every noun in BP consists of
a stem, which may optionally be followed by one ~r more derivative suffixes,
which must be followed by the gender suffix. In (2) we illustrate an
underived and a derived noun together with their respective morphological
structures.
(2) a. [kAza] 'house'
kaz + a
STEM GENDER SUFFIX
b. [kazerG.] 'housekeeper'
kaz + er + \1
STEM SUFFIX GENDER SUFFIX
(2a) illustrates the morphological structure of the noun [kAza] 'house', where <
the noun stem kaz is followed by the gender suffix g. (2b) shows that the
morphological structure of the noun [kazeru] 'housekeeper' consists of the
noun stem kaz fo?lOWed by the derivative suffix er which is followed by the
gender suffix !i.
3 }n Chapter 3 we will see that the occurrence of certain vowels in BP

depends on their position with respect to primary stress placement.
( The proposal of assuming an obligatory morpheme for gender in BPwas
first presented in CAMARA (1970). Other analysis which developed such a
proposal are discussed in LOPEZ (1979) and SEGUNDO (in preparation) ..
5 The derived form illustrated in (2b) represents an instance of non-
analytic morphology, i.e. the sequence of morphemes is phonologically
interpreted as a single morphological domain. Morphemes may also be grouped
into analytic domains, i.e. where morphemes are phonologically interpreted
within distinct morphological domains. In Chapter 3 we will address the
interface between phonology and morphology where we discuss analytic and non-
analytic morphological domains in BP.
18
We assume that the gender suffix corresponds to a nuclear position which may
either be empty, i.e. 0, or it may be filled with any of the nuclear segments
[i,u,a]. The spelling out of the gender suffix, Le. either as 0 or as
[i,6,aJ, is lexically determined. This is because the same noun stem may have
the gender suffix nuclear position eithfr empty· or filled with any of the
segments [1,6,a], as illustrated in (3).
(3) a. (pas + 0) [pas] i~'
b. (pas + i) [pasi] 'pass"
c. (pas + u) [pasu] 'step'
d. (pas + a) [pasa] 'raisin
Nouns may be either feminine or masculine. We assume that the information
concerning the gender of a noun, i.e. whether it is feminine or masculine, is
determined in the lexicon. This is because no principle rules the information
for gender. That is, it is idiosyncratically defined.
In sum, we assume that the lexicon provides information concerning the gender
of a noun as well as how the gender suffix nuclear position will be spelled
out. In (4) we illustrate feminine and masculine forms where fhe gender suffix
nuclear position is either empty or spelled out by [i,n,al. The parentheses
enclose the morphological structure of the nouns and the forms inside the
square brackets correspond to their respective phonetic realizations. Below
each form we provide the lexical information on g~nder and how the gender
suffix nuclear position 1s spelled out.
(4) a. (kaki + 0) [kaki] 'persimmon'
GENDER: Masculine
GENDER SUFFIX: 0
(pas + 0) [pas] 'peace'
GENDER: Feminine
GENDER SUFFIX: 0
6 Evidence for the proposal that the gender suffix corresponds to a

nuclear positiQP is given in SEGUNDO (in preparation). She proposes that empty
gender suffix nuclear positions and gender suffix nuclear positions which are
spelled out by [1] are both lexically empty. Whether a gender suffix nuclear
position remains empty (cf. (3a» or whether it is spelled out by [iJ (cf.
(3b» follows from the information provided in the lexicon.
7 In (4) we illustrate only final and penultimate stressed forms.
Antepenultimate stressed forms will be considered shortly.
19
·: .
b. (kak + i) [kAki] 'khaki'

GENDER: Masculine
GENDER SUFFIX: i
. (gaz + i) [gazi] 'gauze'
GENDER: Feminine
GENDE~ SUFFIX: i
c. (pat + 11) [pato.] 'duck'

GENDER: Masculine
GENDER SUFFIX: fi
(trib + 0.) [tribo.] 'tribe'
GENDER: Feminine
GENDER SUFFIX: u
d. (kar + a) [kara] 'bloke'

GENDER: Masculine
GENDER SUFFIX: a
(fak + a) [faka] 'knife'
GENDER: Feminine
GENDER SUFFIX: a
(4a-d) illustrate a masculine and a feminine noun respectively. In (4a) the
gender suffix nuclear position is empty. In (4b-d) (he gender suffix nuclear
positions are spelled out by [i,u,a] respectively.
Let us consider how final and penultimate stress patterns are assigned. 9 If
we consider the final stressed form (kaki + 0) [kaki] •persimmon' and the
penultimate stressed form (kak + i) [kaki] 'khaki' we observe that in both
forms primary stress falls on the final nuclear position of the noun stem. If
8 Usually feminine nouns do not have the gender suffix spelled out as
[0.]. Except for the example given in (4c), i.e. (trib + 0.) [tribo.] 'tribe',
we only find shortened forms of some feminine nouns where the gender suffix
is realized as [11]. For example: [m5tu] from [motosiklEta] 'motorcycle' or
[f5to.] from [fotografia] 'photography'. Notice that in the shortened forms the
gender suffix occurs as [0.], whereas in their unshortened counterparts it
occurs as [a].
9 The analysis of final and penultimate stress assignment we present in
the following pages is proposed in SEGUNDO (in preparation). Antepenultimate
stress assignment is addressed later in this section.
20
CHAPTER 2: WORD STRBSS IX BP
we take derived forms into consideration, for example (fak + er + u) [fakeru]

'cutlery set', we observe that primary stress falls on the final nuclear
position of the derivative suffix. We may thus generalize that primary stre~~
falls on the final nuclear position of the noun stem or derivative suffix.
Consider (5). .
(5) / \ / \
s wsw
ORO R0 R 0 R' 0 R
I I I I I I
, I I I I I
N N N N I N
I I I I I I I I I
.••.- ..• _._ _. __.._•. -. ·__··..·-·····I·······-i-·····-I-······,_·······-,·· .- -- _.. _ ··_···-1- ..···--,- .--1- -·-1· .-.. . - ...•_•.•- .._ -_.•.- _ ..- .. _ __.•.•
x x x x x x x x X
I I I I I I I I
I I I I I I I I
k a k i 0 k a k r
[kaki] 'persimmon' [kaki] 'khaki'
(5) shows .that final and penultimate stressed forms present a binary foot
where the strong node is assigned to the final nuclear position of the no~~
stem and the weak node is assigned to the gender suffix nuclear position.
We thus conclude that in final and penultimate stressed forms primary stress
is assigned in like manner, i.e. the strong node is assigned to the final
nuclear position of the noun stem or derivative suffix.
We have seen in (l) that primary stress may be final, penultimate or
antepenultimate. The proposal that primary stress falls on the final nuclear
position of the noun stem or derivative suffix allows us to group final and
penultimate stressed forms into a single binary foot pattern as illustrated
in (5). We now have to account for how antepenultimate stress is assigned.
Recall that nouns in BP consist of a stem optionally followed by a derivative
suffix which is followed by the gender suffix. Therefore, in an
antepenultimate stressed form such as (makin + a) [makina] 'machine' primary
stress falls on the second-to-last nuclear position of the noun stem. In this
case we have a ternary foot as illustrated in (6).
10 The proposal that in final and penultimate forms primary stress falls
onto the final nuclear position of the noun stem or derivative suffix was
first presented in LOPEZ (1979). SEGUNDO (in preparation) develops such a
proposal. Secondary stress assignment and word level stress are addressed in
the followtng section.
11 In derived forms the strong node is assigned to the final nuclear
position of the derivative suffix and the weak node is assigned to the gender
suffix nuclear position. The stress pattern for the derived form (fak + er +
u) [fakerfi] 'cutlery set' is given below:
s/'....:w
f a k e r u
21
CHAPTER 2: WORD STRESS IN SP
~
(6)
s
w w
0 R 0 R 0 R
I.
I
t
I
I
I
I
1
,
I
I
I
I
I N I
I N I
I N
I I I I I I
I I I I I I
X X X X X X
I I .I I I I
I I I I I I
m a k i n a
(6) shows that the strong node is assigned to the second-to-last nuclear
position of the noun stem. The weak node which immediately follows the primary
stressed nucleus is assigned to the final nuclear position of the noun stem.
The final weak node is assigned to the gender suffix nuclear position. We now
have to determine whether when primary stress falls on the second-to-last
nuclear position of a noun stem it is ruled by some principle or whether it
is idiosyncratically determined.
If primary stress is idiosyncratically assigned then we can expect to find
noun stems wi th identical segmental sequences where primary stress wi 11 either
fallon the final nuclear position of the noUn stem, i.e. (CVCVC + V), or
primary stress will fallon the second-to-Iast, nuclear position of the noun
stem, i.e. (CVCVC + V). This is because pri~ary stress may fall either on the
final or on the second-to-Iast nuclear position of a noun stem or derivative
suffix (cf. (5,6)). Thus, if we find noun stems with identical segmental
sequences which differ only with respect to primary stress assignment we
conclude that primary stress is idiosyncratically determined. On the other
hand, if primary stress is ruled by principles and parameters we expect that
these principles and parameters will define primary stress placement.
In BP we do find pairs of forms whose noun sterns present the same segmental
sequence where in one case primary stress falls on the final nuclear position
of the noun stem and in another case primary stre~s falls on the second-to-
last nuclear position of the noun stern. These forms are illustrated in (7).
(7) a. (tonik + u)
[toniku] 'Tonico'
[t6niku] 'tonic'
b. (azi + a)
[azia] 'heartburn'
[azja] 'Asia'
c. (paris + 0)
[paris] tparis'
[paris) 'pairs'
The examples in (7) illustrate the morphological structure of nouns with their
phonetic realizations shown below it. Note that both the final and the second-
to-last nuclear position of a noun stem may bear primary stres~~
22
CHAPTER 2: WORD STRBSS IX BP
Forms in (7) lead us to assume that whether primary stress falls on the final
nuclear position of a noun stem (corresponding to the penultimate stress
pattern) or on the second-to-last nuclear position of a noun stern
(corresponding to the antepenultimate stress pattern) is idiosyncratically
determined. This is because the noun stems illustrated in (7) present the same
segmental sequence differing only with respect to primary stress placement.
However, one could question whether there is any particular difference in the
phonological representation of forms in (7). It has been noticed in the
metrical literature that in some languages primary stress is attracted to a
rime which branches. That is, rimes which present either branching nuclei or
____----hr.anc.h.ingJ-.ime.s. These langyages are assumed to be quanti ty sensitive. In W _
we illustrate the representation of a long vowel, a heavy diphthong and a
branching rime.
(a) a. R b. R c. R
1
I
I
I 1\ -
N N N \
I I I
I \ I \ I \
X X X X X x
I I I I
\/ I I I I
a a j a s
Long vowels and heavy diphthongs present a structure where two nuClear
positions are associated to a single nuclear constituent. In (8a) we
illustrate the representation of a long vowel, i.e. [a:], and in (8b) we
illustrate the representation of a heavy diphthong, i.e. raj]. In (8c) we
illustrate the representation of a branching rime.-
One could suggest that BP is quantity sensitive. Thus, the primary stressed
rime would either present a branching nucleus, i.e. it would be filled with
a heavy diphthong or a long vowel (cf. (Ba,b}), or the primary stressed rime
would be branching (cf. (Be». The proposal that primary stress in BP is
sensitive to branching rimes cannot be sustained. The first argument for
denying that primary stress in BP is sensi t1 ve to branching rimes follows from
the fact that we find penultimate stressed vowels preceded by a branching
rime, e.g. (ahtig + u) [ahtigii] 'article'. If primary stress were sensitive
to branching rimes' we would expect the bra~ching rime to be primarily
stressed, I.e. Jahtigii), what is not the case. ._
However ,one -could claim that in a form such as [ahtigii] -'article' primary
stress falls on the second-to-Iast nuclear position because the primary
stressed vowel corresponds to the representation of a long vowel - which is
a branching structure (cf. (8a». According to this proposal, the form
[ahtigii] 'article' woul have the following representation:
12 Notice that antepenultimate stressed branching rimes may occur in BP,

e.g. (ahtikO] 'arctic'.
23
CHAPTBR 2: WORD STRESS II BP
(9) 0 R 0 R 0 R
I I
1\ I I
I
N \ N I N
\ l\ I
I
X X X X X X X
I I I I I
I 1 I \I I I
a h t i g u
There are two arguments against the proposal that BP has long vowels. The
first one concerns the lack of experimental evidence for assuming that primary
stressed vowels in BP are long. The second argument against the proposal that
BP has long vowels concerns the phonological behaviour of languages which have
_ _Lo.ng_....Y.O.Wal~s-.and-..he..a...'lY---dipA.thQng.s-ln_ suc~g:es there is usuaIU _
phonological contrast between long and short vowels as well as segmental
restrictions concerning the vowels in the vowel-glide sequences which
correspond to heavy diphthongs. More specifically, a reduced set from the
overall set of vowels in the language may fill the vowel position in the
vowel-glide sequences which correspond to heavy diphthongs.
BP does not display the typical behaviour of languages which have long vowels
and heavy diphthongs. That is, there is no contrast between short and long
vowels nei ther ~re there segmental restrictions imposed on vowel-gl ide
sequences in BP. 1 Since there is neither phonetic nor phonological evidence
for assuming that BP has long vowels and heavy diphthongs we have to exclude
the proposal that primary stressed vowels in SP present the structure of a
branching nucleus (cf. (9».
The facts we have discussed above lead us to assume that primary stress in BP
is idiosyncratically assigned. Whether primary stress falls on the final
nuclear position of a noun stem, i.e. (kaki + 0) [kaki] 'persimmon' and (kak
+ i) [kaki] 'khaki', or whether it falls on the second-to-last nuclear
position of a noun stem, i.e. (makin + a) [ma.kina] 'machine', is determined
in the -lexicon. We then conclude that primary stress in BP is lexically
determined where feet may be either binary or ternary. In (10) we present
final, penultimate and antepenultimate stressed forms, their corresponding
morphological structure, and their respective stress pattern.
(10) a. (kaki + 0) [kaki] 'persimmon'
/ \
s w
0 R 0 R 0 R
I I I I I
I I I I I
I I
I N I N N
I I I I I
I 1 I I I
X X X X X
I I I I I
I I I I I
k a k i 0
13 A complete description and analysis of vowel-glide sequences in BP is

provided in Chapter 7.
24
.: .
CHAPTER 2: WORD S!RESS IN BP
~
s
w w
0 R 0 R 0 R
I I I I I
I I I I I
I I
I N N I N
I I I I
I I I I
X X X X X X
I I I r I I
I I I I I I
m a k i n a
In this section we have considered h~w primary stress is assigned. We showed

that primary stress is lexically determined where feet may be either binary
or ternary. In the following section we will consider how secondary stress and
word level stress are assigned.
2.2. Secondary Stress and Word Level Stress'

In this section we wi 11 be concerned wi th the parameters involved in the
construction of metrical trees. More specifically, we will' consider the
analysis proposed by SEGUNDO (in preparation) which defines the parameters
involved in secondary stress assignment and word level stress. Our proposal
d~ffers from SEGUNDO's in two major aspects. First, we assume that primary
stress assignment is lexically defined, whereas she claims that primary stress
is parametrically determined. Whereas her proposal for stress assignment of.
final and penultimate stressed forms provides a generalization regarding
primary stress assignment (cf. section 2.1.) she does not account for stress
assignment of antepenultimate stressed forms.
The second difference from SEGUNDOS' s proposal concerns secondary stress
assignment and the construction of word level trees. Whereas she claims that
metrical structure is built onto lexical representations we argue that
metrical trees are built onto the projection of nuclear heads after
phonological processes apply. We assume that only the nuclear. heads which are
not governed are projected to the level of nuclear projection, which we will
represent as R. The metrical structure, i.e. secondary stress and word trees,
is thus built onto the level of nuclear projection. Evidence for our proposal
25
.: .
CHAPTER 2: WORD STRESS IX BP
concerning the construction of the metrical structure, i.e. that only those
nuclear heads which are not governed are projected to the level of nuclear
projection where metrical structure is built, is given in Chapter 5 where we
address governing relations between adjacent nuclear posi tions. Metrical trees
are built as follows:
(11)
a.Primary stress is assigned at the level of lexical
representation where feet may be either binary or ternary.
b.Phonological processes apply and those nuclear heads which are
not governed are projected to the level of nuclear projection.
c.Secondary stress is assigned onto the level of nuclear
projection according to the following parameters:
- feet are binary and left-headed.
- foot construction begins from the right edge.
d.Right-headed word trees are built onto the projection of
metrical heads.
According to the parameters defined in (11) forms such as (parana + 0)
[parana] 'Parana' ; (paralEl + a) [paralEla] 'parallel' : (satElit + i)
[satEliU] 'satellite'; have the following metrical trees:
(12) a. (parana + 0) [parana] •Parana'
w--------.. s word tree level
:-=z:::::::: w
sI I
~w
I I
foot level
~ ~ ~ ~
R R R R nuclear projection
I I I I
I I I
0
I
R
I
0
I
RI
0
I
R
I
0 R
I
I I
N I Ii I N
I I
N
II I
I
I
I
I
1
II I
I
I
I
X X X X X X X
I I I I I I I
I I
I
P a
I
r
I
a n a
I
"0
b. (paralEl +[paralEla] 'parallel'
(1)
w~s word tree level
F=:::::::
sr w :----
sI W foot level
I I
~ ~ ~ ~
I I I I
I I I I
0 R
I
0
I
R
I
0
I
R
I
0
I
R
I
I
1 I I
I
I
I
N I N I N I N
II I
I
II I
I
II I
I
II I
I
X X X X X X X X
I I I I I I I I
I
I
P
I
a r
I
a
I
i E
I
i a
26
·: .
CHAPTER 2:- WORD STRESS IN BP
c. (satElit + i) [satEliti] 'satellite'

w~s word tree level
l~
s
I
sI wI w
I
foot level
• + + ±
R R RI ~nuclear projection
I I ,
, I I ,
ORO
I I I
R
I
0I R
I
0, R
I
I I I I I I I I
II ~I II ~I I
I
~I II ~I
X X X X X X X X
I I I I I I , I
I
S
,
a tl i J. t' J.
I
Ell
(12a-c) illustrate the metrical trees of final, penultimate and

antepenultimate stressed forms respectively. The phonological representation
of forms in (12) presents onset-rime sequences where both constituents are
associated to a single skeletal position which is filled with segmental
material. That is, neither branching constituents nor constituents whose
skeletal positions are empty or constituents which are not associated to a
skeletal position occurs. Nevertheless, the phonological representation of
constituents plays a role with respect to stress . assignment. More
specifically, there are metrical constraints . imposed on posttonic
constituents. Such constraints will be stated in the following section.
2.3-. Metrical Constraints Imposed on COnstituents

In this section we will present the inventory of BP phonological constituents
and we will state the metrical constraints imposed on posttonic constituents.
Consider (13).
(13) a. 0 b. 0 c. 0 d. 0
I
I
I
1\
I
I I \
X X X X
I I I
I / \ I I
a a ~ a 13
e. R f. R g. R 0
I
I
I
I
I 1\ I
N N
, N\ I
I
I I
I I \ I
X
I
I
,
X
I
,
X
I
X
I
I
X
a 0 a 13 or
(13a-d) illustrate the inventory of BP onsets and (Be-g) illustrate the
inventory of BP rimes. Let us consider each representation in (13) in more
detail.
27
CHAPTER 2: WORD STRESS 1M BP
(13a) illustrates an onset position which is associated to a single segment.

An onset position which is associated to a single segment in BP maybe filled
wit h a n y 0 f the !~o 1 low i n g s e 9 men t s : [p, t , k, b,
d,g,m,n,ft,l,A,f,v,s,z,~,~,r,h].
(13b) illustrates an onset position filled with a complex consonant. In the

representation of a complex consonant we have two segments associated to a
single onset position. T~e complex consonants [kw, gw] may be associated to
an onset position in BP.
(l3c) illustrates the representation of a branching onset. It represents a
constituent governing domain where the far left position is the head and the
far right position is the complement. Branching onsets present a stop or a
non-sibilant fricative in the head position and a liquid in the complement
position. The following branching 16 onsets occur in BP:
[pr,tr,kr,br,dr,gr,vr,fr,pl,tl,kl,bl,gl,fl].
(13d) illustrates an onset constituent which is not associated to a skeletal
position. In Chapter 5 w will consider the role of onsets which dominate no
skeletal position in BP.17 .
(13e) iUustrates a nuclear position which is associated to a single
segment. A nuclear position which is associated to a single segment in BP
may be filled with any of the following segments: [a,e,€,i,o,~,u,l,u,aJ.19
14 For the internal representation of segments which may occupy an onset

position see KLV (1990), HARRIS (1990) and YOSHIDA (1991).
15 In Chapter 6 we will consider in detail the role of complex consonants
in BP. See also SILVA (1992).
16 [vI] occurs only in few loan words, ~e.g. [v1adimih] 'Vladimir'. *[d1]
does not occur.
17 For the role of onsets which dominate no skeletal position see also
CHARETTE (1988, 1991).
18 A single nuclear position may also have two segments associated to it.
In this case we have
-
the representation of a light diphthong, i.e. R
I
I
N
I
/)
Light diphthongs will be discussed in Chapter 5 where we address governing
relations between adjacent nuclear positions.
19 See Chapter 1 for the internal representation of nuclear segments in
BP. In Chapter 3 we will consider the distribution of BP nuclear segments in
relation to the primary stressed position.
28
""
(13f) illustrates an empty nuclear position. Empty nuclear positions are

subject to a special type of government called proper government. For details
on proper government see KAYE (1990a) and CHARETTE (1988, 1991). Preliminary
analyses of empty nuclear positions in BP are presented in SEGUNDO (1989) and
SILVA (1989c). CAVACO (in preparation) analyses empty nuclear positions in
European Portuguese.
(13g) illustrates the representation of a branching rime. Rimal positions are
governed by the following onset position (cf. KAYE (1989C».20Rimal positions
in BP may be filled with the following segments: [s,h,l,n].
Given"the inventory of -SPphono1ogicalcOIIStituents-' 'in (l-31 let'us-cons'ider
the metrical constraints imposed on posttonic constituents. Consider (14).
(14) The posttonic rime in antepenultimate
stressed forms may not be branching, i.e.
*[hekuhsu] but [heklihsu] 'recourse'
*{petisku] but [petisku] 'snack'
(14) determines that whenever the final rime of a noun stem is branching, e.g.
(hekuhs + 0), primary stress will fall 2yn the n?clear position of the
branching rime, i.e. [hekuhsu] 'recourse' .
. (15) The second posttonic, onset in antepenultimate
stressed forms may neither be filled with the complex consonants
[kw, gw], nor with the BP palatal consonants, i.e. [fi,l,S,~], or
with the segment {h].
*[inikwa] but [inikwa] 'iniquitous'
* [galina] but [galida] 'hen'
*[manila] but [manila] 'shackle'
*[b6liSi] but [boliSi] 'bowling'
*[kuru~a] but [kurli~a] 'owl'
*[sigahti] . but (sigahu] 'cigarette'
(15) determines that whenever the final onset of a noun stem is filled with
any of the segments [kw,gw,n,l,S,~,h] primary stress will fallon the final
nuclear PO~ition of the noun stem, e.g. (boliS + i) [boliSi] 'bowling' but
It [bOHSi].
20 When the segment !! occurs in a rimal posi tion it triggers the

nasalization of the vowel in the preceding nuclear position~ e.g. (sant + u)
[satu] 'saint'. In the cases the segment! occupies a rimal position it is
vocalized, e.g. (alt + ti) [awtu] 'high'. The segment h may also be manifested
as a velar fricative, i.e. [X], or as a trill, i.e. [r], or as a retroflex,
Le. {t]' when it occurs in a rimal position, depending on a variety of
sociolinguistic factors (cf. OLIVEIRA (1983».
21 This constraint was first observed by HARRIS (1983) on his analysis
of Spanish.
22 HARRIS (1983) observed that palatal consonants cannot occupy the final
onset position in antepenultimate stressed forms in Spanish.
29
.: .
In (14-15) we have stated metrical constraints imposed on posttonic

const! tuents. Some other metrical constraints are imposed on forms which
present prevocalic and postvocalic g11<1es, Le. [jV, wV] and [Vj, Vw). For
example, in forms presenting posttonic prevocalic glides primary stress must
fall on the nuclear position prece<1ing the glide-vowel sequence, Le.
[familja] 'family' but *[familja]. Also vowel-glide sequences are not allowed
to occur in posttonic position, i.e. *[patew), t[patewta] or *[patatew). In
Chapters 6 to 8 we discuss the relationship between the metrical structure and
forms which present glide-vowel sequences and vowel-glide sequences in BP. We
will claim that glides in BP are derived from a nuclear position filled with
a high vowel. Evidence for our claim is supported by the behaviour of the
metrical structure in forms which present prevocalic and postvocalic glides.
We have seen that primary stress in BP may be either final, penultimate or
antepenultimate (cf. (1». That is, primary stress may fall at most on the
third-to-Iast nuclear position of a noun. There are however some exceptional
forms where primary stress may fallon the fourth-to-last nuclear position,
e.g. [tEkinika) 'technic and [hitimika] 'rhythmic'. For reasons which are not
I
yet clear these forms present the suffiX -ik.

In this section we presented the inventory of BP phonological const! tuents and
we have stated metrical constraints imposed on posttonic constituents.
2.4. Conclusion
In this Chapter we have considered how word stress is assigned in BP. We
showed that primary stress is lexically determined where feet may be either
binary or ternary. We have also addressed the construction of metrical trees.
We argued that metrical trees are built onto the level of nuclear projection.
Finally we have stated some metrical constraints imposed on posttonic
I
constituents. In the following Chapter we will discuss the distribution of BP

vowels in relation to the primary stressed position.
30
~CKAPTER 3: TKE VOWEL SYSTKK OF BP
CHAPTER 3
THE VOWEL SYSTEM OF BRAZILIAN PORTUGUESE
3.0. Introduction
In this Chapter we will see that the occurrence of certain vowels in BP
depends on their position with respect to the primary stressed vowel. That is,
~ome vQwels only occur 1n posttonic position, others only in primary stressed
position whereas others may occur either pretonically or in primary stressed
position. We propose that the distribution of BP oral vowels with respect to
the primary stressed position is accounted for by charm constraints imposed
on nuclear segments. In this thesis we are concerned with phonological
processes involving sequences of adjacent oral vowels. Once the distribution
of vowels is defined with respect to the primary stressed position we will be
able to establish which sequences of vowels are found in BP. Thus, we can
analyse phonological processes involving sequences of adjacent oral vowels.
3.1. Primary Stressed Vowels

We have seen in the previous Chapter that primary stress in BP may be either
final, penultimate or antepenultimate. In {l} we reproduce forms from Chapter
2 which illustrate these three patterns.
(1) i- Fi·nal stress

a. [sabia] 'Sabia (bird)'
b. [oboE] 'oboe'
c. [dosie] 'dossier'
d. [ikarai] , Icarai (beach) I
e. [faraS] 'pharaoh'
f. [kra6] 'Krao (tribe)'
g. [bali] 'trunk'
ii. Penultimate stress
a. [piada] , joke'
b. [duElii] 'duel'
c. [kue).ii] 'rabbit (masc) ,
d. [saida] 'exit'
e. [viSla] 'viola'
f. [di6gii] 'Diogo'
g. [saudi] 'health'
ii1. Antepenultimate stress
a. [mfikina 1 'machine'
b. [elisi] 'helix'
c. [ezitti] 'success'
d. [sHaba] 'syllable'
e. [kaStlkii] 'chaotic'
f. [kr6nikii] 'chronic'
g. [uvUla] 'uvula'
31
CHAPTER 3: rHE VOWEL SYSTEM OF BP
Forms in (1) show that in primary stressed position the following oral vowels
occur: [a,e,E,i,o,~,u). The internal representation of these vowels is given
in (2).
(2) f
(vo .A+ or A+ = [a)
{.+ .(A .1.0)0)+ = [e]
(At .1°)0 = [E]
f
(.+ • 1°
(.+.(A .~)o)+
= [i]
(A+ .!f)o
= [0]
= [~]
(.+ .~)+ = [u]
Notice that- all the--segments in (2) p.~t-lea.st one posit-i-¥e1¥--c-hame4--
element in their internal representation. Indeed this fact reflects a
requirement imposed on the phonological system of BP, according to which a
segment occurring in primary stressed pesi tion must have at least one
positively charmed element in its internal representation. Accordingly, we do
not find the lax hi~h vowels [1] and [u] - whose internal representations are
(vo.lo)o and (vo.!f) respectively - in primary stressed position since these
segments do not present any positively charmed element in their internal
representation. With respect to the schwa - which contains the positively
charmed element A+ in its internal representation, Le. (A+ .yo)o - it is never
found in primary stressed position. Given that the segment [a] has the
posi ti vely charmed element At in its internal representation we have to
explain why it does not occur in primary stressed position. According to our
proposal, a segment which presents at least one positively charmed element in
its internal representation should be able to occur in primary stressed
position. The fact that a schwa is never primary stressed follows from .RIa
contrainte de Kryocephalie" proposed in LOWENSTAMM (1986) rhiCh aims to
account for the fact that a schwa is never long, i.e. *[a:]. According to
this constraint, a segment which has the cold element as its head cannot be
the head of a branching nucleus, i. e. « N•
1\
x x
\/
a
Let us extend LOWENSTAMM's proposal as follows: a segment which has the cold
element as its head cannot be the head of a domain. Such a proposal accounts
for the fact that a nuclear position filled with the segment [a] ~annot bear
primary stress. This is because a nuclear position which bears primary stress
is. the head of the metrical domain. Thus it cannot be filled with a segment
such as [a] which has the cold element as its head. In the following section
we consider the distribution of posttonic vowels.
1 ftLa contrainte de kryocephalie" states that a cold-headed segment, such

as a schwa, cannot be the head of a branching nucleus. It has been proposed
to account for stress phenomena in Arabic. More specifically, it accounts for
the fact that a schwa is never stressed in Arabic. For a detailed discussion
on this issue see LOWENSTAMM (1986).
32
CKAPfER 3: raE VOWEL SYSTEM OF BP
3.2. Posttonic Vowels .

In this section we consider the vowels which occur following primary stress.
Considering the penultimate and antepenultimate stressed nouns in (1) one
observes that the vowels occurring after the primary stressed posi tion can be:
[u,i,a]. In (3) we illustrate the internal representation of these vowels.
(3) [ii]
[i]
[a]
The occurrence of [ii,i,a] following the primary stressed vowel in BP reflects
a requirement imposed on the BP phonological system, according to which
posttonic vowels must be formed by a single element and the resulting segment
must be neutrally charmed. The single elements At, UO and 10 are selected. The
elements 10 and u<> correspond to the neutrally charmed lax vowels [i] and [ii]
respectively. However, the element At corresponds to the vowel [a] which is
a positively charmed segment. We have seen in Chapter 1 that the internal
representation of a segment such as [a] can be either At or (vD .~t)t. We assume
that in order to satisfy the requirement that a neutral segment must occur
after primary stress, the elements which form the segment [a], i.e. (VO.~t~O,
are flipped around, yielding (At .yO)o. The phonetic manifestation of (At.y )0
is a schwa, i.e. [a], which is a neutrally charmed segment.
We have seen that posttonic vowels must be neutrally charmed segments. It
follows from this constraint that nasal vowels should not occur in posttonic
position. This is because nasal vowels are positively charmed segments. Indeed
this is the case. It is important to mention that the constraint which
prevents nasal vowels from occurring posttonically (because posttonic nuclear
segments must be charmless) operates in colloquial speech. However, in careful
speech, posttonic nasal vowels may occur.; For example, forms like [via~ej]
'trip', rima] 'magnet' and [f:>rli] 'forum' which present posttonic nasal vowels
in careful speech are pronounced as [via~i], rima] and [f~ru] in colloquial
speech. That is, posttonic nasal vowels do not occur, as predicted by charm
theory.
In like manner, positively charmed oral vowels, i.e. [a,e,i,o,u], occur in
posttonic position in careful speech, e.g. [silaba] 'syllable'; [nadegas]
'buttocks'; {abitu] 'habit'; [atomu] 'atom'. However, in colloquial speech
posttonic lax oral vowels are realized instead, Le. [silaba]; [nadigas];
[abitu]; [ t i t W n G . ] . · .
In this section we have discussed the distribution of posttonic vowels. In the

following section we consider the distribution of pretonic vowels.
2 The phonological representation of these forms are respectively:

/via~En0/, /iman0/ and /f::>rfin0/. They represent cases of vowel-nasal consonant
sequences where the nasal consonant does not surface (cf. Chapter 1).
33
CHAPTER 3: THE VOWEL SYSTEK OF BP
3.3. Pretonic Vowels

Let us now consider the vowels occurring in pretonic position. Consider the
forms in (4).
(4) a. [orista] 'hourly worker'
b. [duelista] 'duellist'
c. [:;,ritia] 'hour (diminutive)'
d. [duElisimu] 'the best duel'
(4) shows that in pratonic position either tense or lax mid vowels occur.
--Neverthal-es-s-;-f"orms----irr-(-4-)--a-re---tnstance-s-ot-dertvad-noun-s-.-{--4-a)--ancr(-4-c-)--are--------
derived from the noun stem :;,r- 'hour' and (4b) and (4d) are derived from the
noun stem duEl- 'duel'. Evidence that the noun stem vowel of derived forms in
(4) presents a lax mid vowel comes from the underived forms of these nouns
when the noun stem vowel occurs in primary stressed position, Le. [Sra]
'hour' and [duElu] 'duel'.
What is interesting about the forms in (4) is that in the derived forms in
(4a-b) the noun stem vowel is phonetically manifested as a tense mid vowel,
1.e. [e] and [0], whereas in the derived forms in (4c-d) the noun stem vowel
is phonetically manifested as a lax mid vowel, i.e. [E] and [:;,]. Consider~ng
forms like (4), we have to explain why noun stems presenting an underlying lax
mid vowel require some of their derived forms to present a tense mid vowel in
pretonic position (cf. (4a-b» whereas in other derived forms a pretonic lax
mid vowel occurs (cf. (4c-d». -
One way of -accounting for the alternations between tense and lax vowels in
derived forms such as in (4) is to assume that the occurrence of pretonic
tense and lax mid vowels is unpredictable. That is, speakers have to learn
which derived forms present a pretonic lax mid vowel and which ones present
a pretonic tense mid vowel. If the occurrence of pretonic lax and tense mid
_vowels is unpredictable, we should find tense and lax pretonic mid vowels not
only in derived nouns - as illustrated 1n (4) - but also in underived ones.
If pretonic tense and lax mid vowels occur at random, nothing should prevent
underived nouns from presenting both tense and lax pretonic mid vowels.
Consider forms in (5) which are instances of underived nouns.
(5) a. [venezuEla] 'Venezuel~'
b. [pare6] 'Pareo' _
c. [teoria] 'theory'
d. [poEta] 'poet'
e. [proeza] 'prowess'
f. [marionEti) •puppet ,
(5) shows that in underived nouns only ~se mid vowels occur in pretonic
posi tion. It is important to say that there is no phonological contrast
between tense and lax mid vowels in pretonic position, whereas the contrast
between tense and lax mid vowels is observed in primary stressed position as
34
CHAPTER 3: THE VOWEL SYSTEK OF BP
illustrated in (6) • 3
(6) a. [sed!] 'thirst'
b. [sEd!] 'headquarters'
c. [f6hma] 'baking pan'
d. [f~hma] 'form'
We have seen that in underived nouns pretonic mid vowels must be tense (cf.
(5)). On the other hand, in derived nouns either tense or lax pretcinic mid
vowels occur (cf. (4}). The fact that pretonic lax mid vowels only occur in
derived forms (cf. (4c-d» - but never in underived ones (cf. (5» - leads us
to posit that the morphological structure should be relevant in accounting for
forms which present pretonic lax mid vowels. We will propose that derived
forms which present pretonic lax mid vowels, e.g. [~rifta] and [duElisimu] (cf.
(4c-d», present a different morphological structure from derived forms in
which a tense mid vowel occurs pretonically, e.g. [orista] and [duelista] (cf.
(4a-b».
Let us assume that lax mid vowels only occur in primary stressed position in
BP (where indeed phonological contrast is observed (cf . (6) ) • If our
hypothesis is correct, we have to assume that in derived forms such as [~rina]
and [duElisimu] (cf. (4c-d» there must have been a point within the
derivational history of these forms where the lax mid vowels occurred in
primary stressed position. For reasons which we will consider shortly,
throughout the derivational process it happened· that the lax mid vowels
surfaced in a position which is no longer primary stressed. We propose that
the morphological structure of forms in (4c-d) is.
(7) a. « in + a)
~r+a) [~rina]
b. «duEl+u) isim + u) [duElisimu]
~
We propose that the parentheses in (7) enclose independent morphological

domains, i.e. a morphological domain which corresponds to what is called a
word. Forms in (7) present two sets of parentheses. The innermost parenthesis
enclose an underived noun, i.e. a noun stem followed by the gender suffix, and
the outermost parenthesis enclose the innermost parenthesis and a derivational
suffix followed by the gender suffix. The morphological categories enclosed
in each set of parenthesis in (7) represent a morphological domain whica
corresponds to a cycle. The phonological interpretation of forms like (7)
follows from'the innermost cycle to the outermost one. We will consider next
the derivation of [~rifia] (cf. (7a)) assuming that any form presenting a
pretonic lax mid vowel is derived in like manner.
The innermost cycle - which corresponds to the underived noun [~ra] - is
phonologically interpreted when the lax mid vowel occurs in stressefi position.
3 In some northern dialects of BP, e.g. Recife (in the state of

Pernambuco), pretonic nuclear positions may present lax mid vowels, e.g.
[p~Eta] 'poet' and [parE6] 'Pareo'. However, no phonological contrast between
pretonic tense and lax mid vowels is observed in any dialect of BP.
35
CHAPTER 3: THB VOWEL SYSfKM OF BP
(a) ./'...
s w
0 R 0 R 0 R 0 R
I
I
I
I
,
I
I
I
I
I
I
I
I
N I N N 1
I N
I I I I I I
I I I I I I
X X X X X X
I I I I 1 I
I I I I I I
« :::l r a) i 11 a
The innermost cycle is then accessible to the outermost cycle.

(9) ~ ~w
s w s
0 R 0 R 0 R 0 R
I I I I I I
I I I I I I
N I
1 N N 1
I N
r I I I I I
I I J I I I
X X X X X X
I I I I I I
I I I I I I
(:) r a i fi a)
Notice that in (9) there are two strictly adjacent nuclear positions. That is,
the initial nuclear position of the suffix is strictly adjacent to the final
nuclear position of the noun. We claim that a sequence of strictly adjacent
nuclei is the adequate environment for vowel deletion to take place. More
specifically, we propose that a nuclear point is deleted when it is followed
by a strictly adjacentnucleus. 4 It follows from our proposal that the final
nuclear position of the noun in (9) is deleted since it is followed by a
strictly adjacent nucleus. This process is illustrated in (10).
(10)
s
/'.w s
A w
0 R 0 it 0 R 0 R
I
I
I
I I I
I
I
I ,
I
I I
N
I
I
I
I
I
)(
I
N
,
I
I
N
I
I
X x 1- x x X
I I I I
I
I I I I I I
(:::l r a 1 fi a)
The deletion of the nuclear position in (10) yields a structure which presents
a sequence of onsets. According to the theory every onset must be followed by
a nucleus (cf. Chapter 1). Therefore, a structure presenting two consecutive
onsets is ill-formed since one of the onsets is not preceded by a nucleus. We
propose that the pointless onset in (10) is deleted in order to allow an
onset-nucleus sequence to occur. 'rfius, the final onset from the noun is
4 We assume that vowel deletion - or more precisely the loss of a nuclear

posi tion - takes place as the consequence of the establishment of right-headed
internuclear governing relations which are considered in details in Chapter
5.
36
CHAPTER 3: THE VOWEL SYSTEM OF BP
followed by the initial nucleus from the suffix. Nuclear heads are projected
and the metrical structure is constructed. The final derivation of the form
[~rifia] is illustrated in (II).
(ll) ~
w s word tree level
r---
ssw foot level
I I I
.!. .L .!.
R R R nuclear projection
I I I
I I I
-------·--------------O--R---o-R--o--R- ---- --_..---._--:-------------------.-.-------
I I I I I
I I I I I
N I
I N I
I N
I I I I
I I I I
X X X X X
I I I I I
I I I I I
(~ r i fl. a)
domain, i.e. a word, they will not present lax mid vowels in pretonic position
(since any single domain in BP does not allow lax mid vowels to occur
pretonically). Given that forms in (l2) must present tense mid vowels, we have
evidence that (12) contains a single domain. The forms [orista] and [due1ista]
are obtained.
According to the analysis we have proposed in the preceding pages we should
find mainly two ways of grouping morphemes in BP. In one case morphemes would
be grouped as «x)y), Le. where more than one morphological domain is
involved. In the other case the morphemes are yrouped as (xy) when they are
analysed as a single morphological domain. The proposal to analyse
morphological domains as above is presented in KAYE (1989d). According to KAYE
····an--t-nstance- --trr---wh1-ctr·--morpho-togrca1:·-cat:egorteg--are---group-eu---crs·_-(-(-xly-)-or-
«x)(y» is called 'analytic' morphology and an instance in which
morphological categories are grouped as (xy) is called 'non-analytic'
morphology. In (13) we illustrate the forms which involve analytic and non-
analytic morphology that we have considered in the preceding pages.
(l3)
a. « ::Ir+Q) iii + a ) [::Irida] 'hour (dim)'
b. «duEl+u) iS1m + u) [duElisimu] 'duel (sup) I
c. (or + ist + a) [orista] 'hourly worker'

d. (duel + ist + a) [duelista] 'duellist'
(13a-b) illustrate instances of analytic morphology which involve internal
domains. (13c-d) illustrate instances of non-analytic morphology which involve
single domains. We agree with HALLE (1986:6) who claims that whether or not
an affix is CYClic, or analytic in our terms, is an idiosyncratic and variable
property of it. It follows from this that the suffixes -iii (cf. (13a» and -
1s1m (cf. (13b» require analytic morphology whereas the suffix -ist (cf.
(13c-d» requires non-analytic morphology. Most of the suffixes in BP require
non-analytic morphology. That is, suffixes are added directly to noun stems
and they are phonologically interpreted as a single morphological domain. I~
(14) we illustrate some suffixes from BP which require analytic morphology.
5 In the previous pages we considered cases in which more than one

morphological domain is analysed as «x)y). Morphological domains may also be
grouped as «x)(y». Such cases are discu~sed in KAYE (1989d).
6 flI shall assUme that whether or not an· affix is cyclic is not a
property of the morphological rule by which it is assigned, but is rather an
idiosyncratic and variable property of the affix." HALLE (1986:6).
7 Indeed we propose that the phonological representation of the suffixes
-iii and -1s1m is: ~
o NON 0 NON 0 N
I I I I I
I I I I I
X X X X X X X X
I I I I
I I I I
Z i ii Z i s i m
Notice that in the representations above the suffixes are preceded by the
38
CHAPTER 3: THB VOWEL SYSTEK OF BP
(14) a. -ift 'diminutive'

b. -isim 'superlative'
c. -m~t 'adverb'
Evidence that the suffixes illustrated above require analytic morphology comes
froll other phonological processes which, like the al temations between lax and
tense vowels we have described before, take place when a vowel is primary
stressed, or more precisely is within the dominant fo~t of a domain. We will
discuss next the process of vowel nasalization in BP.
It has been observed that a primary stressed vowel in BP must be nasalized
wnerr it ts--foUowe-ci-- by--a- lra-sal---consonant -tcL- --PONTES (-1972-Y,--VANDRESSEN-
(1975), SHAW (1986». On the other hand, a pretonic vowel which is followed
by a nasal consonant can be either nasal or oral. These facts are illustrated
in (15).
(15) a. [oselinu] *[oseanu] 'ocean'
b. [p~na] *[pe.na] 'pity'
c. {final * [fina] 'thin'
d. (~uaneti ] " [~uClneti] 'bunion'
e. [diamliti ] " [diClmliti] 'diamond'
f. [dionizju] " [dionizjll] 'Dionisio'
Forms in (lSa-c) show that a primary stressed vowel followed by a nasal
consonant must be nasalized. Forms (lSd-f) show that a pretonic vowel followed
by a nasal consonant might or might not be nasalized. The alternation between
oral and nasal vowels in pretonic position (cf. (lSd-f» leads us to expect
any pretonic vowel followed by a nasal consonant to be either nasal or oral.
Nevertheless, some forms require a pretonic vowel followed by a nasal
consonant to be nasalized. Consider forms in (16).
(16) . a. [osednif'1u] *[oseaninu] 'ocean (dim)'
b. [panisima] * [pen1sima] 'most pitiful'
c. [finam~ti] *[finam~tI] 'thinly'
Forms in (16) present suffixes which, according to our analysis, require
analytic morphology (cL (l4)}. Given that the suffixes in (l6) require
analytic morphology these forms will present the following mprphological
structure.
segment !. We assume that ! is a floating consonant which does or does not

attach the initial onset position of the suffix according to government
constraints. Assuming the representations above we account for variable forms
presenting diminutive and superlative suffixes, i.e. [~ina] " [:::>razifta] 'hour
(dim)' and [m:::>lisimft] " [m:::>lizisirnu] 'softest'. For details on the analysis
of these forms see SILVA (l989a).
8 Further evidence that the suffixes illustrated in (14) require analytic
morphology comes from the process involving breaking up of branching onsets
which is considered in Chapter 6. We refer the reader to SILVA (1989b) for
details on this issue.
39
CHAPTER 3: THB VOWEL SYSTEM OF BP
(17) a. «osEan+u) in + Q) [osednifiu]

b. «pEn+a) 1sim + a) [penisima]
c. «fin+a) met + i) [fInameti]
Let us consider the derivation of [penislma] (cf. (17b». We assume that any
form which requires a pretonic vowel followed by a nasal consonant to be
nasalized is derived in like manner. Consider (18).
(18) /"-
ws
( ( pen a ) i s i m a)
·-TfieulIi-nermos-t-cycl:e-rs·pnonoIoglcaI1Y-fiiterpretea--fn-(fSTWlien-Ule··noun····rp-enci]-·
'pity' is derived. The noun stem vowel must be nasalized since it occurs in
primary stressed position followed by a nasal consonant within the domain. We
then proceed to the outermost cycle as illustrated in (19).
(19)
~ ~
s w s· w w
pen a i s i m a)
In (19) there is a sequence of adjacent vowels which corresponds to a sequence
of strict1y,adjacent nuclear positions. That is, the initial nuclear position
of the suffix is strictly adjacent to the final nuclear positlon of the noun
stem. We have already proposed that in a sequence of adjacent nuclear
positions the far left position is delete9 (cf. (10». It follows froni· our
proposal that the final vowel of the noun in (19) is deleted. This proc~ss is
illustrated in (20).
(20)
~
ssw w
(p e n ~ i s i m a)
The final derivation of the form [penisima] is illustrated in (21).
(21) /'..
w s word tree level
I s~
s w w foot level
pen i s i m a)
In a form like [penisima] apretonic vowel followed by a nasal consonant must
be nasalized, i.e. *[pEnisima] (cf. (l6b». Vowel nasalization is obligatory
in this case because the noun stem vowel is nasalized in the innermost cycle.
Vowel nasalization must OEcur when a primary stressed vowel is followed by a
nasal consonant (cf. (18». Consider the derived forms in (22).
(22)
a. [osean~grafu] - [osean~grafu] 'oceanographer'
b. [penozu] - [penozu] 'distressing'
c. [fineza] - [fIneza] 'soft manner'
40
Forms in (22) show that in derived forms a pretonic vowel followed by a nasal
consonant can be either oral or nasal. The alternation between oral and nasal
vowels in forms in (22) is accounted for by the non-analytic nature of the
derivat!ve suffixes. The morphological structures of forms in (22) are
(23) a. (osEan + ~raf + 6)
b. (pEn + oz + u)
c. (fin + ez + a)
Forms in (23) are phonologically interpreted as a single morphological domain.
Therefore, as in underived nouns (cf. (l5d-f» a pretonic vowel followed by
a nasal consonant in (23) can be either oral or nasal. In the preceding pages
we have considered phonological processes which involve analytic and non-
analytic morphological domains. We have shown that pretonic lax mid vowels,
e.g. [~rifta] and [duElisimu], and pretonic nasal vowels, e.g. [osedniftu] and
[penisima], must occur when the analytic suffixes -in-, -isim- and -met- are
involved in derivational processes. Throughout this thesis - unless otherwise
stated - derived forms that we will consider involve non-analytic morphology,
i.e. they are phonologically interpreted as single morphological domains. Let
us now return to the discussion of constraints imposed on pretonic vowels in
BP. Underived forms and forms involving non-analytic morphology must present
pretonic positively charmed oralYowels, i.e. [a,e,i,o,u]. In (24) we
illustrThte the internal representation of BP positively charmed oral
vowels.
(24) (vo. A+)+ or A+ = [a]
( ...+. (A+ .1°)0)+ = [e]
( ...+.1°)+ = [i]
("'+. (A+ .!f)o)+ = [0]
( ...+.!f)+ = [u]
9 The proposal to analyse morphological domains we have discussed in the

pre'cedin-g pages- is 'presenfed in -de-taBs-in KAYE' (l9S9df and-kAYE- '-WRGNAtiD
(1990). It is important to say that within this view a given suffix may
require analytic morphology in one dialect and non-analytic morphology in
another dialect. It is also possible that a suffix which gefterally requires
analytic morphology may require non-analytic morphology when it occurs with
a restricted set of noun stems. Both of these possibilities are attested in
BP. That is, suffixes which require analytic morphology in the Bela Horizonte .
dialect (cf. (14» require non-analytic morphology in the Natal dialect (cf.
SEGUNDO (in preparation». It also happens that the analytic suffixes
presented in (14) occur in a non-analytic manner w~h a restricted set of noun
stems. For a fuller description on analytic and non-analytic morphological
domains in BP see SILVA (1989a).
10 Nasal vowels - which are also positively charmed segments - may occur
in pretonic position of underived forms and forms involving non-analytic
morphology. Forms involving analytic morphology may present complex charmless
segments in pretonic position, e.g. [~rifia] 'hour (diminutive)' and
[duElisimu] 'the best duel'.
41
In this section we have discussed the distribution of pretonic vowels in BP.

In the following section we summarize the discussion of the BP oral vowel
system we presented in the preceding pages and we define which sequences of
adjacent oral vowels we find in BP.
3.4. Summary of the SP Vowel System

In (25) we present the distribution of BP oral ·vowels together with the charm
requirements which must be satisfied with respect to the primary stressed
position. .
(25)
a. Primary stressed position: The segments must have at least one
posi ti vely charmed element in thei r internal representation.
[a,e,E,i,o,::l,u]
b. Following primary stress: The segments must be formed by a single
element and the resulting segment must be neutrally charmed. [u,i,a]
c. Preceding primary stress: The segments must be positively charmed.
[a,e,i,o,u]
At this point we are able to define which sequences of adjacent oral vowels
we shall find in BP. Any primary stressed vowel, i.e. [a,E,e,i,o,::l,u], can
potentially be followed by [u, i, a] , and any. primary. stressed vowel can
potentially be preceded by [a,e,i,o,u]. We also expect to find pretonic
sequences of vowels presenting members of the set [a,e,i,o,u] and sequences
of posttonic vowels presenting members of the set [u,i,a].
Throughout this thesis we will be dealing with phonological processes which
involve sequences of adjacent nuclear positions. More specifically, we will
be investigating sequences of adjacent nuclear positions where one of the
nuclear positions is filled with a high vowel and the other nuclear position
is filled with some other vowel. In order to analyse such processes we have
to define the underlying representation of high vowels in BP. This is the
topic of lhe following sectn~rf~-----_···-- ..... -_. .-.-- .----- -----
3.5. The Underlying Representation of High Vowels

In this section we will argue that the underlying representation of high
vowels in BP corresponds to the lax high vowels [I] and [u] - whose internal
representations are 1° and !f respectively. Whether a tense or a lax high
vowel is phonetically manifested depends on charm constraints imposed on
nuclear segments.
4#"
Evidence for such a proposal - of assuming that the underlying representation

of high vowels in BP corresponds to lax high vowels - comes from nasal vowels.
We have seen that nasal vowels are derived from a fusion operation in which
the positively charmed nasal element Nt combines with a charmless segment (cf.
Chapter l). The fact that the nasal element only combines wi th cha,t;J!lless
segments follows from the Charm Principle which establishes t~~t expressions
42
CHAPTER 3: THE VOVEL SYSTEM OF BP
with like charm cannot combine. Given that the nasal element is positively
charmed it cannot combine with positively charmed elements. If high vowels in
BP corresponded to the tense vowels Iii and lui - which represent positively
charmed segments, Le. (..+.1°)+ and (...+.yll)+ - we would not expect high vowels
to be nasalized. This is because the positively charmed segments Iii and lui
cannot combine with the positively charmed nasal element N+. Indeed, high
vowels may be nasalized in BP, e.g. [st] 'yes' and[u] 'one'. The fact that
high vowels may be nasalized in BP, provides us with evidence that' the
underlying represennation of high vowels in BP corresponds to the lax high
vowels IiI and luI.!
According to our proposal, the underlying representation of high vowels in BP
corresponds to lax high vowels. We have seen that both tense and lax high
vowels occur in BP, e.g. [piada] 'joke' and [Elisi] 'helix' (cf. (1». Thus,
we have to explain the conditions under which a tense or a lax high vowel is
phonetically manifested. We propose that whether a tense or a lax high vowel
is realized depends on charm constraints imposed on nuclear segments,
according to their position in relation to the primary stressed position.
Let us first consider primary stressed vowels. Recall that a nuclear position
which bears primary stress must be filled with a segment which presents at
least one positively charmed element in its internal representation (cf.
(2Sa». The lax high vowels [I] and [u] - whose internal representations
consist of the charmless elements 1° and UO - do not present any positively'
charmed element in their internal representations. Then, their occurrence in
primary stressed posi tion violates charm constraints. We propose that in order
to satisfy charm requirements 'the lax high vowels [1J and [u] receive either
the ATR element ...+ or the nasal element ~ to their internal representations.
The positively charmed segments [i,i,u,6], which present a positively charmed
element in their internal representations, are realized. Thus, charm
requirements imposed on primary stressed nuclear heads are satisfied.
Let us now consider pretonic high vowels. We have seen that pretonic nuclear
positions must be filled with a positively charmed segment {cf. (2Sc». We
assume that in order to satisfy charm requirements imposed on pretonic nuclear
heads either the ATR element or the nasal element ~ which transmits positive
charm to a phonological expression - is added to the internal representation
of the lax high vowels Iii and lui. Tense high vowels, i.e. [i,u], or nasal
high vowels, i.e. [I,Q] - which are positively charmed segments - will occur
in pretonic poSitiO~i Thus, charm requirements imposed on pretonic nuclear
heads are satisfied.
11 KAYE's (198ge) discussion of the distribution of nasal vowels in

French observes that high vowe~ cannot be nasalized, i.e . • [i], *[u] and
*[y]. Interestingly, his analysis shows that the underlying representation of
high vowels in French corresponds to tense high vowels.
12 Throughout this thesis we will be discussing phonological phenomena
which involve BP oral vowels. Thus, we will restrict our analysis of charm
requirements to cases in which the ATR element is added to the internal
representation of a high vowel so that a tense oral high vowel, i.e. [i,u],
is realized. Cases in which the nasal element is added to the internal
43
CRAPTER 3; fRE VOWEL SYSTEM OF BP
Regarding posttonic nuclear positions we have seen that they must be filled
with a neutrally charmed segment (cf. (25b}). Thus, the lax high vowels [i]
and [6] - which are neutrally charmed segments - satisfy charm requirements
imposed on posttonic nuclear positions.
3.6. Conclusion
In this Chapter wehave seen that the occurrence of certain vowels in BP is
dependent on charm constraints imposed on nuclear heads. Such constraints
define which sequences of adjacent vowels we shall find in BP (cf. (25}).
Throughout this thesis we will be investigating forms which present sequences
of adjacent nuclear positions. OUr analysis aims to account for the processes
of vowel coalescence in BP with emphasis on the process of vowel-glide
alternations. In the following Chapter we present a review of the literature
concerning high vowels and glides in BP. We will see that further research is
still needed on this topic. Then, in the remaining Chapters of this thesis,
we will present our analysis to account for the processes of vowel-glide
alternations in BP.
representation of a segment in order to satisfy charm requirements are not

analysed in this thesis because they would require a detailed discussion of
processes involving vowel nasalization in BP which would lead us beyond the
scope of our
research.
44
CHAPTER 4
REVIEW OF THE LITERATURE
4.0. Introduction
A perusal of the phonological literature on BP shows that there has been
controversy as to whether high vowels and glides are both underlying segments
or if the latter are derived from the former. In this Chapter we will consider
both views in different frameworks. First we consider CAMARA's (1953,1970)
phonemic 'ana-i ell --are'
ys1s. 'Later'we"conside"i --works wh'i based-- on
the -linear'
generative framework. Finally, we discuss BISOL's (1989) analysis which aims
to define the syllabic structure of diphthongs in BP wi thin a non-linear
framework.
4.1. A Phonemic View .,-. ,- - .-
4.1.1. Glides as Underlying Segments

The first Brazilian linguist - CAMARA - proposed in 1953 that glides should
be analysed as independent phonemes in BP. 1 Such a proposa1,was based on the
interpretation of glides as consonantal segments within the syllabic system
of BP. Analysing glides as consonantal segments allows one to fit them into
the closed syllable pattern CVC. This type of syllable would occur in forms
such as [mes] 'month' and [pasta] 'suitcase' as well as in forms like [paw]
'wood' and [gajta] 'tin fife'. The basic argument for assuming this proposal
is that it provides a simpler phonotatic system. This is because, if glides
were analysed as vowels, a syllable of the type CVV would have to be assumed.
This type of syllable would occur only in forms presenting vowel-glide
sequences, e.g. [paw] 'wood' and [gAjta] 'tin fife'. In order to provide a
simpler syllabic system - which excludes CVV syllables - glides were then
analysed as consonantal segments.
Although the proposal of analysing glides as consonants provides a simpler
syllabic system - which excludes CVV syllables - it implies a more complex
phonemic inventory - which includes glides. On the other hand, if glides are
analysed as vowels, a simpler phohemic ipventory - which excludes glides - is
provided, and a more complex syllabic system - which includes CVV syllables -
must be assumed. At this stage it appeared that the interpretation of glides
as either vowels or consonants depended on simplifying either the syllabic
system or the phonemic inventory of BP. Evidence about either assumption was
lacking.
-4'
,In 1970 CAMARA rejected his initial proposal that glides behave as consonants
fitting into the closed syllable pattern CVC. He claimed then that glides have
to be analysed as vowels - fi tUng into a CVV syllable type. CAMARA's
argument for assuming that glides must be interpreted as vowels concerns the
1 As we will see shortly, this proposal was later rejected by CAMARA

(1970).
45
.: .
CHAPTER 4: RBVIEW OF THE LITBRATURE
distribution of "R" sounds in Portuguese.

Two types of "R" are assumed in Portuguese. The so-called bland R - which is
phonetically manifested as a tap - and the so-called strong! - whose phonetic
manifestation varies considerably from dialect to dialect. Contrast between
bland and strong Rs is found only in intervocalic position, e.g. [karu] 'dear'
and [kahii] I car'. Other than intervocalically the bland R occurs only
following an obstruent, e.g. [pratii] 'plate'. The strong R, besides
intervocalically, occurs in every other environment: that is, either in
initial position, e.g. [hatu] 'rat'; or in final position, e.g. [mAh] 'sea';
or closing a syllable, e.g. [kahta] 'letter'; or following a closed syllable,
e.g. [ishaEw] 'Israel'.
As CAMARA observes, if glides are interpreted as consonantal segments -
fitting into the closed syllable pattern CVC - we should find the strong R
after a glide. This is because the strong R follows a consonant which occurs
in a closed syllable, cf. [ishaEw], i.e. *[israEw]. Nevertheless, we do not
find the strong R after a glide but instead th bland R occurs following it,
r
e.g. (plewra] 'pleura' and [~ajIii] 'Jairo'. The fact that the bland R
follows a glide supports CAMARA's claim that glides have to be interpreted as
vowels. This is because bland Rs occur after a vowel, e.g. [karul 'dear', but
not after a closed syllable, *[israEw].
Given that wi thin the syllabic system glides behave like vowels, CAMARA (1970)
proposes that glides are "positional variants· of the high vowels Iii and lu/ ..
In other words, he proposes that glides are derived from high vowels. In the
following ,section we discuss his proposal:
4.1.2. Glides Derived from High Vowels

CAMARA (1970) proposes that glides are "positional variants" of the high
vowels /i/ and /u/, Le. he claims that glides and high vowels are in
complementary distribution. According to his proposal high vowels are
manifested as their corresponding glides when a high vowel is flanked by a
stressed vowel, e.g. [gajta] 'tin fife' and [kwAdrii] 'picture', i.e. *[gaita]
and *[kuAdru]. In other environments, i.e. when a high vowel is' flanked by a
vowel which does not bear primary stress, high vowels are in free variation
with glides, e.g. [kaipira] - [kajpira] 'bumpkin' and [~uazeru] - [~wazeru]
,Jt..'Clzei ro' .
2 We will transcribe the strong R as [h]. It may also be phonetically

manifested as a trill [f], or as a velar fricative [X], or as a retroflex It]
depending on a variety of sociolinguistic factors. For a description and a
sociolinguistic analysis of "R" sounds in BP see OLIVEIRA (1983). For further
detail on the distribution of "R" sounds in BP see CAMARA (1970), LOPEZ
(1979), CAGLIARI (1982), MAlA (1981), SHAW (1986). .
3 A problem with CAMARA's proposal is that he does not explain why strong
R's cannot follow a glide, e.g. *[plewha]. Given that strong R's may follow
a vowel, e.g. [kahu] 'car', we should find strong R's following a glide since
according to his proposal glides behave like vowe~s.
46
CHAPTER 4: REVIEW OF TBE LITERATURE
Nevertheless, CAMARA's proposal encounters some difficulties. The first one

is related to his claim that high vowels flanked by a primary stressed vowel
are manifested as a glide, e.g. [kw&drft] but *[ku&drft). It follows from his
proposal that we should not find forms which present a high vowel surrounded
by a primary stressed vowel. This is because, according to him, any high vowel
surrounded by a primary stressed vowel should be manifested as a glide.
Nevertheless, in forms such [kwa).ii) ~[kua).ii) 'curd' -either a glide or a high
vowel may be phonetically manifested in a position which immediately precedes
a primary stressed vowel. According to CAMARA, a pronunciation such as [kua).u]
would be impossible because in this form the high vowel ~ immediately precedes
a primary stressed vowel. Thus, CAMARA's proposal that glides occur whenever
------------a high -vowel is adjacent t40 a primary -stressed -vowel -does not -account for-arf
forms in BP, cf. [kualii].
Another problem with CAMARA's analysis involves forms in which glides occur
in posttonic position, e.g. [agwa) 'water'. According to CAMARA a high vowel
flanked by a vowel which does not bear primary stress is in free variation
with its corresponding glide. It follows from this that a form like [&gwa] -
where a glide is -followed by a vowel which does not bear primary stress -
should alternate with a form like [&gii.a] - where a high vowel is pronounced.
Nevertheless, this is not the case. A form like [agwa) 'water' cannot be
pronounced as *[agiia], i.e. with a posttonic sequence of vowels. According to
CAMARA both [agwa] and *[agiia) should be possible pron~ciations for 'water'.
we have seen that CAMARA's analysis is problematic in accounting for some
forms in BP (cf. [kualii] ~ [kwa'\U] and-[agwa)) • However, he raises some
interesting points. The first is the evidence he provides for the fact that
glides must be interpreted as vowels (cf. the discussion on "R II sounds above).
The second interesting point he makes is that the role of the primary stressed
vowel is fundamental in accounting for when a glide or a high vowel occurs in
BP. In the following section we will discuss how glides and high vowels have
been analysed within the linear generative framework.
4.2. A Linear Generative View

4.2.0. Introduction
Most of the phonological analysis on BP are based on the linear generative
framework (cf. HEAD (l964), NARO· .(1971), ISTRE (1971), JEROSLOW (1974),
REDENBARGER (1976), LOPEZ (1979), BEDDOR (1982), SHAW (1986»). Regarding the
phonological interpretation of glides, there has been controversy on whether
glides are derived from high vowels or whether glides are underlying present
in phonological representations. In this section we will see that either
assuming that gli~s are underlyingly present or that glides are derived from
their corresponding high vowels is problematic within the linear generative
4 In Chapter 6 we consider in detail forms such as [kwadrii] 'picture' -

but not *[kuadriiJ, Le. where the glide does not alternate with its
corresponding glide; and forms such as [kwa,\ii] ~ [kualii] 'curd', i.e. where
either a glide or a high vowel may be phonetically manifested.
47
framework. 5
4.2.1. Glides as Underlying Segments

In this section we consider the proposal that glides are underlyingly present
in the phonological system of BP. Within the linear generative framework the
difference between high vowels and glides depends on whether a [high] segment
is specified as either +. or - [syllabic]. In {Ii we illustrate the
phonological representation of the segments [i,j,u,w] .
._.J.!1__ .li.1. . . . -.1jJ. . .__ .______ ._ . _. .. _
+ syllabic -- SYllabiC]
[+ high ] [+ high
[u] [w]
+ syllabic
[ + high
J r- SYllabiC]
L+- high
Notice that the segments [i,j] and [u,w] differ only with.respect to the
feature [syllabic]. Whereas the segments [i,u} represent a syllabic peak the
segments [j,w] do not. Thus, the segments [i,u] may bear stress whereas the
segments [j,w] may not.
A problem with the assumption that glides are underlying present in BP comes
from the distribution of posttonic glides in relation to primary stress.
Primary stress in BP may be ei ther final, penultimate or antepenultimate, e. g.
[para] 'Para', [kaza] 'house' and [silaba] 'syllable' (cf. Chapter 2). If we
assume that glides are underlying segments in BP then we expect to find
antepenultimate stressed forms where a glide follows primary stress. This is
because glides do not count as a syllable peak. Thus, nothing should prevent
stress from falling on the third-to-last syllable peak, i.e. *[gajtika],
*[familja] and *[g6jaba], should be possible forms. However, this is not the
case. Thus, the proposal that glides are underlying segment~ in BP does not
explain why stress cannot fallon the third-to-last syllable peak in forms
where a glide follows primary stress .
. In this section we have seen that the proposal that glides are underlying
segments in BP does not account for the lack of antepenultimate stressed forms
5 We have opted for not discussing any specific work which is based on
the linear generative framework. This is because such works alm to provide a
general analysis of the phonological system of BP rather than to provide an
analysis which deals with a specific topic. Regarding the phonological
behaviour of glides and high vowels in BP the authors we have quoted above
only state their assumption as to whether glides are underlyingly present or
derived from their corresponding high vowels - without presenting arguments.
6 In (I) we only specify the relevant features for the representations
of these segments.
48
.: -
CHAPTBR 4: REVIEW OF THE LITBRATURE
where a glide follows primary stress. In the following section we consider the
proposal that glides are derived from high vowels.
4.2.2. Glides Derived from High Vowels

Recall that what distinguishes a high vowel from ". ,s corresponding glide is
the feature [syllabic] (cf. (1». Thus, the assumpt- ';1 that glides are derived
from their corresponding high vowels leads us to ~ ~it a rule which turns a
[+syllabic] segment into its [-syllabic] counterp~:. Let us first consider
the process of deriving prevocalic glides. Considei (2).
(2) thigh I ----> [-syllabic] / [tsyllabic]
!-stress
Rule (2) states that an unstressed hig? vowel turns into its corresponding
glide when it is followed by a vowel. We have seen that whether a glide
alternates or fails to alternate with its corresponding high vowel depends on
its position-with respect to the primary stressed vowel. That is, in pretonic
position either a glide or a high vowel may be manif~-sted ([piada] - [pjada]
Ijoke ') whereas in posttonic position only a glide: ;- y occur ([sabja] 'wise'
but * [sabial).
In order to express the relationship between thf: " _ :.:i tion in which the high
vowel occurs with respect to the primary stre,c -vowel, Le. whether it
occurs pretonically or posttonically, -one would 1-,- - i to stipulate awkward and
ad hoc constraints to rule (2) : This is because wi t ;~ ... n the formalism available
in the linear generative framework, a vowel or a [+syllabic] segment is
defined as either [+stress] or [-stress]. Thus, to determine whether an
unstressed vowel occurs pretonically or posttonically would require the
stipulation of awkward and ad hoc constraints. Let us now consider the process
of deriving postvocalic glides. Consider (3).
(3)lthi gh
-stress
I---) [-syllabic] / [tsyllabic] ----
Rule (3) states that an unstressed high vowel turns into its corresponding
glide when it is preceded by a vowel. In a similar way to the proposal of
deriving prevocalic gl~des (cf. (2»1 a problem which arises from deriVing
postvocalic glides from a rule such as (3) is that it would be awkward and ad
hoc to express the fact that posttonic postvocalic glides do not alternate
with their corresponding high vowel ([gajta] 'tin fife' but *[gaita]) whereas
pretonic postvocalic glides do ([kajpira] - [kaip1ra] 'bumpkin'). In other
words, once again the relationship between the position in which the high
vowel occurs with respect to the primary stressed vowel, i.e. whether it is
pretonic or~sttonic, requires the stipulation of awkward and ad hoc
constraints to rule (3).
7 Rule (2) must define that the high vowel does not bear primary stress.
This is because, unlike unstressed vowels, a primary stressed high vowel
cannot turn into its corresponding glide, i.e. [pia] 'sink' but *[pja] or
[asai] rA~ai' but *[asaj].
49
There are further problems wi th analyses based on the linear generative

framework. These problems involve the theoretical assumptions of the model
which allows one to express more phonological processes than those found in
natural languages as well as in its assumptions a phonological rule does not
address explicitly the relationship between an event and the environment in
which it takes place.
For example, rules (2) and (3) determine that a high vowel becomes [-syllabic]
when it is flanked by another vowel, i.e. a [+syllabic) segment. Taking the
processes illustrated in (2) and (3) into consideration, one could ask some
of the following questions: Could a high vowel turn into its corresponding
glide when it is flanked by a consonant~ Why is it that unstressed vowels are
SUbject to the process of gliding and not stressed ones~ Why are high vowels
subject to the loss of their syllabicity when flanked by another vowel~ Could
glides turn into their corresponding high vowels?
Without going into theoretical details of the model, let us address some of
these questions. For example: could a high vowel turn into its corresponding
glide when it is flanked by a consonant? In the formalism available for the
linear framework it is possible to formalize processes such as those
illustrated in (2) and (3) - according to which a high vowel turns into its
corresponding glide when it is flanked by a vowel. It.is also possible to
formalize a process according to which a high vowel turns into its
corresponding glide when it is flanked by a consonant, i.e. [+consonantal)
segment. However, a phonological process which turns high vowels into their
corresponding glides before or after a consonant does not occur in natural
l.anguages. Thus, a linear framework allows one to express phonological
processes which are recurrent in natural languages as much as phonological
processes which do not occur at all.
Let us now address the question of why only unstressed vowels may turn into
a glide unlike stressed ones. The fact that only unstressed vowels may turn
into a glide 1s seen as a stipUlation without any connection to the stress
system of the language. Nevertheless, as we have been discussing throughout
this Chapter, the processes involving high vowel-glide alternations in BP are
closely related to the stress system of the language. However, no connection
between the process" of gliding and the stress system "of BP is established
within the linear framework.
In the previous pages we have addressed some problems in analysing
phonological phenomena related to high vowel-glide alternations in BP within"
a linear framework. We have seen that the problems met by such analyses are
related to the fact that in a linear framework stress is viewed as a segmental
feature attributed to individual segments rather than as a property of a
domain. We have also seen that in the formalism available in a linear
framework one may express more phonological processes than those which occur
in natural languages. Finally, in such framework one does not express the
relationship between a phonological event and the environment in which it
occurs. In the following section we consider how glides and high vowels have
been analysed within a non-linear framework.
.to ". _
50
.: .
4.3. The Non-linear View

4.3.0. Introduction
Wi thin non-linear frameworks a great deal of work has been done on the process
of vowel-glide alternations (cf. STERIADE (1984), KAYE AND LOWENSTAMM (1984),
KENSTOWICZ AND RUBACH (1987), MAROTA (1988), CARREIRA (1988), BOOIJ (1989».
These works aim to define the phonological representations of falling and
rising diphthongs in different languages.
Few works on BP are based on non-linear frameworks. In this section we discuss
-·----Bl-sot"""'s[I9"8"9) worK wniCh------':o our-Knowledge lstne on1r attempt to investigate
the phonological representation of diphthongs in BP.
4.3.1. Bisolls Analysis

The hierarchical structure of phonological representations assumed by B1S0L
(1989) is illustrated in (4).
(4) • syllable tier
/ \
0 R rime tier
I
I
X x prosodic tier
I I
I I
[features] [features] melodic tier
She assumes that the so-called rising diphthongs - which consist of a glide
followed by a vowel, e.g. [kjabu] 'okra' - are derived from a sequence of
rimes. The so-called falling diphthongs - which consist of a vowel followed
by a glide, e.g. [pawza] 'pause' - are assumed to be either lexically present
or derived. Let us first consider her analysis of falling diphthongs whose
phonological representations may be either (Sa) or (5b).
(5) a. R b. R
I \ I
I
X X X
I
- I
. [ ... ] [ ... ] ..
I I
[ ... ] / [ \... ]
[pawza] 'pause' [f~jra] 'market'
[hejnu] 'kingdom' [pejSi] 'fish'
[sEw] 'sky' [k6wru] 'leather'
..,.
8 This work reflects the research done during her sabbatical year at
Stanford University (1988). In her bibliographical references no other works
on the same topic in BP are mentioned.
51
: .
According to BISOL (Sa) represents the structure of a heavy diphthong and (5b)
represents the structure of a light diphthong. 9 Heavy diphthongs correspond
to "true" or phonological diphthongs and are associated to two skeletal
positions. Light diphthongs are derived from assimilation processes at t~~
melodic tier (cf. (4» and are associated to a single skeletal position.
According to BISOL's proposal heavy diphthongs differ from light diphthongs
in the sense that the former are phonologically present and the latter are
derived. Another difference between heavy and light diphthongs is that the
former cannot be reduced whereas the latter are likely to be reduced.
BISOL assumes that the same diphthong, e.g. [ej], might have two different
phonological representations. That is, a diphthong such as [ej] may have the
structure of a heavy diphthong in a formiilCe [h~jniir-'-Klngdom'--~-rsar);;:----------------
which, according to her, is a phonological diphthong - and the diphthong [ej]
may have the structure of a light diphthong in a form like [fejra] 'market'
(cf. (5b» - which is a diphthong derived at the melodic tier. It follows from
BISOL's analysis that in a form like [hejnu] 'kingdom' the diphthong [ej] is
not subject to reduction - since it corresponds to a heavy diphthong (cf.
(Sa» - whereas in a form like [fejra] 'market' the diphthong [ej) may be
reduced to [e], since it corresponds to a light diphthong (cf. (5b».
BISOL's analysis aims to explain wny some falling diphthongs are reduced, Le.
[ow] is reduced to [01 and [ej] is reduced to reI, whereas others are not. In
order to solve this problem she proposes two different structures shown in
(5). Diphthongs having a structure such as (5a) cannot be reduced whereas
these presenting the structure (5b) are subject to reduction.
Let us .consider in detail what has been called the process of diphthong
reduction in BP (cf. NUNES (1951), NARO (1973), ALVARENGA ET AL (1989».
Basically two diphthongs are said to undergo a reduction process: [ow] is
reduced to [0] and [ej] is reduced to reI. That is, forms such as [k6wru]
'leather' and [fejra] 'market' may be manifested as [k6ro] and [feral
respectively.
According to BISOL's analysis the phonological representation of diphthongs
which may undergo reduction corresponds to the structure (5b), i.e. a light
diphthong. As she notes, the assUmption that diphthongs which are subject to
reduction are derived and therefore have thelftructure of a light diphthong
(cf. (5b» is problematic (our translation):
9 GP also assumes the notions of heavy and light diphthongs. However, as

we will see later (Chapters 5 to 8) the notions of light and heavy diphthongs
assumed in GP differ considerably from those assumed by BISOL.
-
10 Unfortunately, BISOL neither illustrates how the process of deriving
light diphthongs operates nor gives references to any work which would
exemplify such phenomenon.
11 "A analise correta desses dados nao esta totalmente esclarecida.
Apresentaremos duas linhas de argumentos, ambas com certes problemas. Todavia
os dois tratamentos sugerem que 0 ditongo em pauta nao existe na estrutura
profunda de itens lexicais".
52
CHAPTER 4: REVIEW->- OF THE LIfERATURE
"The correct analysis of these data (diphthongs which may be reduced)

is still not totally clear. We will present two different analyses,
both of which reveal some problems. Nevertheless, both treatments
suggest that .the diphthongs under investigation do not exist in the
underlying representation of lexical items." (BISOL: 193)
BISOL's proposal that diphthongs which are subject to reduction are derived
is indeed problematic. Contrary to BISOL we assume that diphthong reduction
in BP reflects a lexical reanalysis process rather than a phonological one.
Evidence that diphthong reduction reflects a lexical reanalysis process comes
from hypercorrection which occurs in informal dialects of BP. That is, -0- and
---------e----torrns-whrc-h----are-not-htst-ortcal-d·i-p·hthongs-;-e~[_pe_ra-]-'_pEra:r-'-and-[_p6vft-]--
'people', are (hyper)corrected to -ow- and -ej- as in [pejra] and [p6wvft] in
informal dialects of BP.
Let us now consider BISOL's proposal that the phonological representation of
heavy diphthongs corresponds to the structure illustrated in (Sa), i.e. two
skeletal positions are associated to the same nuclear const! tuent. Recall that
she claims that heavy diphthongs are underlyingly present. Two problems arise
from the claim that heavy diphthongs are lexically present in BP. The first
concerns the phonological behaviour of heavy diphthongs with respect to
primary stress. Primary stress in BP may be either final, penultimate or
antepenultimate, e.g. [para] 'Para', [kaza] 'house', [silaba] •syllable , (cf.
Chapter 2). If we assume that heavy diphthongs Fepresent two skeletal
positions associated to the same nuclear constituent, then we predict the
existence of forms which present antepenultimate stressed heavy diphthongs,
e.g. [gajtikfil. This is because a heavy diphthong would metrically count as
a single nuclear head. Thus, a form such as [gajtikfi] would represent an
instance of antepen~ttimate stress. However, forms such as *[gajtiku] usually
do not occur in BP.
The second problem with the assumption that heavy diphthongs are leXically
present in BP comes from derived forms which end in a heavy diphthong, e.g.
[muzew] 'museum', and their derived forms, e.g. [muze~lfigfi] 'specialist in
museums'. Recall that the final vowel in penult!mate and antepenultimate
stressed nouns in BP corresponds to the gender suffix which may be spelled out
either as [a,u,i], ego [kaza] 'house', [pulft] 'jump', [tristi] 'sad' (Chapter
2). In derived forms the gender suffix is suppressed when a derivative suffix
is added to the noun stem, e.g. [kazerfi] 'house keeper', [pul&dfi] 'jumped' and
[tristeza] 'sadness'.
If we consider the derived form [muze~lfigU] 'specialist in museums' - which
is derived from the noun [muzew] 'museum' - we notice that the final glide is
suppressed when the suffix -~lugil is added to the noun stem muze-. The fact
that the glide does not occur in a derived form such as Imuze~lfigU] gives us
12 There are few forms in BP which present antepenultimate stressed heavy

diphthongs, e.g. [nawtikfi] 'nautical' and [fahrnasewtikfi] 'pharmaceutical'. We
will show in Chapter 7 that in these forms the heavy diphthong is not
lexically present.
53
.: .
evidence that it corresponds to the gender suffix. l3

.In the preceding pages we have considered BISOL' s analysis of falling
diphthongs. We have seen that her analysis is problematic in the sense that
it assumes two distinct phonological structures for the same segmental
sequence (cf. (5». Furthermore, her assumption that vowel-glide sequences are
lexically present in BP is problematic because she can nei ther account for the
lack of forms which present antepenultimately stressed heavy diphthongs, i.e.
*[gajtiku], nor for the fact that the glide is suppressed in derived forms,
cf. [musew], [muzeSlugu].
Let us now consider BISOL' s proposal to analyse the so-called rising

_·-atpttt11o-ng-S-~-She--p-r-oposes---e-tta:~--gl-nlFVOWel---s-e-quen~crre-11er1VeCl.-rrom-a""'-
sequence of rimes where the high vowel delinks from its rimal position and
either spreads into the preceding onset position (ef. (6a» or into the
following rime (ct. (6b».
(6) a. C V V (C) b. C V V (C)
\1 1/
[+ high] [+ high]
[- stress] [- stress]
According to BISOL the spreading of the high vowel into the preceding onset
(cf. (6a» has priority over the spreading of the high vowel into the
following rime.(ct. (6b». Consider (7).
(7) a. 0 R 0 R b. 0 R
I I I
1\ I I I 1\ 11\
k i a b u k r i 0 w
CCVCV CCVVC
[kjabu] 'okra' [krj6w] 'created'
Notice that BISOL assumes two different derivations for a single process, 1. e.
the derivation of prevocalic glides. In one case she assumes that the high
vowel delinks from "i ts rimal position spreading into its preceding onset, e. g.
[kjabo.] 'okra I (cf. (7a», and in another case the high vowel delinks from its
rimal position spreading into its following rimal position, e.g. [krj6w]
'created' (ef. (7b». Although BISOL proposes the derivations illustrated in
(7) to derive prevoealic glides, she does not give any argument for her claim
that the spreading of the high vowel into its preceding onset (cf. (7a» has
priori ty over the spreading of the high vowel into its following rimal
position (cf. (7b». In fact, to assume two different derivations for a single
process, i.e. the derivation of prevocalic glides, is ad hoc since its aim is
to express the same phonological process.
13 An example in which the gender suffix occurs as a vowel rather than

as a glide is the form [futuru] 'future'. When the suffix -~lagU is added to
the noun stem futur- the gender suffix -u is omitted, i.e. [futur:Slugfi]
'fortune teller'.
54
~CBAPTER 4: REVIEW OF THE LITERATURE
4.4. Conclusion
In this Chapter we have considered how the process of vowel-glide alternations
and the" phonological representation of diphthongs have been treated within
different frameworks. We have seen that these works encounter problems in
explaining when a high vowel alternates or fails to alternate with its
corresponding glide, and in determining the phonological representation of
diphthongs in SP.
In the following Chapter we will discuss governing relations between adjacent
nuclear positions. The proposal on governing relations presented in the
following Chapter will be applied to our analysis of vowel-glide alternations
in SP which is presented in the remaining Chapters of this thesis.
55
CHAPTER 5: ON GOVERNING RELATIONS BETWEEN iUCLEAR POSITIONS
CHAPTER 5
ON GOVERNING RELATIONS BETWEEN
NUCLEAR POSITIONS
5.0. Introduction
In this Chapter we will consider governing relations involving a sequence of
strictly adjacent nuclear positions. It is important to mention that no major
work has dealt specifically with this topic. Therefore, the discussion
addressed in this Chapter aims to provide a general discussion on this issue
as well as to discuss the theoretical implications of governing relations
between nuclear positions within the GP framework. The proposal presented in
this Chapter will be applied to our analysis of vowel-glide alternations in
BP which 1s presented in the remaining Chapters of this thesis. Consider (1).
(1) 0 R 0 'R
I l
I I
N N
x, x
I
I I
VI V2
(I) illustrates a sequence of adjacent positions which are associated to
phonological constituents that belong to the same category, namely nuclei.
Given that the adjacent positions in (I) are associated to constituents that
belong to the same category (nuclei) we have a sequence of identical
phonological units. It has been argued in the Ii terature that a representation
which presents a sequence of identical phonological units is not well-formed.
This constraint has been defined as the Obligatory Contour Principle or the
OCP. According to this constraint, sequences of identical phonological units
are prohibited in phonological representations. Thus, in order to prevent a
sequence of identical phonological units from occurring, phonological
processes apply. In the following pages we will discuss the notion of OCP and
we will address how governing relations involving a sequence of adjacent
nuclear positions have been analysed within the GP framework.
5.1. The Obligatory Contour Principle
In this section we will discuss the notion of the Obligatory Contour Principle
. (henceforth OCP) since this concept has been applied to the analysis of
governing relations between adjacent nuclear positions which we will discuss
later in this Chapter. It has been argued in the literature that a sequence
of identical phonological units is not a well-formed representation. This
constraint is stated in (2).
(2) Obligatory Contour Principle
Sequences of adjacent identical units are prohibited in
phonological representations.
OCP was initially proposed by LEBEN (1973) to account for tone simplification
56
CHAPTER 5: ON GOVERNING RELATIONS BETWEEN NUCLBAR POSITIONS
in Mende. l Consider (3).
(3) a. HL
mbu 'owl'
. b. LH
mba 'rice'
c. HL H HL H
mbu i ---> mbu i 'the owl'
d. LH L LH L
mba ngaa ---> mba ngaa 'rice(pl)'
e. HL L H L
mbu ngaa ---> mbu ngaa 'owls'
f. LH H L H
mba i ---> mba 1 'the rice'
(3a-b) illustrate nouns which present a falling and a rising tone
respectively. In (3c-f) the nouns given in (3a-b) are followed by a suffix.
In (3c-d) the nouns are followed by a suffix whose tone is different from the
adjacent tone of the preceding noun. The tonal pattern of forms in (3c-d)
corresponds to' the tone of the noun followed by the tone of the suffix.
In (3e-f) the nouns given in (3a-b) are followed by a suffix whose tone is
identical to the adjacent tone of the preceding noun. In this case the tonal
pattern of forms in (3e-£) is simplified. That is, the final tone of the noun
which precedes the suffix is deleted. This process isfllustrated in (4).
(4) a. HL L ---> H¥: L ---> H L
mbu ngaa mbu· ngaa mbu ngaa
b. LH H ---> L~ H ---> L H
mba i mba i mba i
(4) shows that in a sequence of identical tonal units the far left tone is
deleted. The process of tone simplification illustrated in (4) is accounted
for by OCP which prohibits sequences of identical phonological units.
The notion of OCP discussed above has been applied to analyses of governing
relations between adjacent nuclear positions. We have seen in (1) that a
sequence of adjacent nuclear positions represents a sequence of identical
1 GOLDSMITH (1990) provides a general discussion of the OCP, addressing

various theoretical issues related to different assumptions of the OCP. For
further details on the notion of OCP see also GOLDSMITH (1919), McCARTHY
(1986), YIP (1988), ODDEN (1988).
51
CHAPTER 5: ON GOVERRIXG RELATIONS BETWEEK NUCLEAR POSITIONS
phonological units (because it presents a sequence of skeletal positions which

are associated to phonological constituents belonging to the same category,
i.e. nuclei). Therefore, we expect a sequence of adjacent nuclear positions
to be simplified due to the fact that it is a sequence of identical
phonological units, which is prohibited under the OCP. In the following
section we will consider how governing relations between adjacent nuclear
positions have been analysed within the GP framework.
5.2. Previous Proposals on Governing Relations Between AdjaCent Nuclear

Positions
5.2.0. Introduction
Recall that government is defined as a binary, asymmetric relationship between
adjacent skeletal positions (ct. Chapter 1). The establishment of a governing
relation defines a governing domain where one of the positions is the head (or
the governor) and the other position is the complement (or the governee). In
a governing domain the head position is filled with a segment which has the
adequate charm value to govern the segment in the complement position. In (5)
we summarize the governing properties of segments (ct. Chapter 1).
(5)a.Charmed segments may govern; charmless segments may be governed.
b.Charmed segments may not be governed.
c.Charm1ess segments may govern if they have a complexity greater
than their governee.
The governing properties of segments stated in (5) define the potential
governing relations between a sequence of adjacent nuclear positions:
(6 ) 0 R o ~ 'R
I
I •
N
I
h
I
I I
X X
I I
I I
VI V2
a. If VI is a charmed segment and V2 is a charmless segment, a

left-headed governing domai~ is established.
b. If VI is a charmless segment which has a complexity greater
than V2, a left-headed governing domain is established •.
c. If V2 is a charmed segment and VI is a charmless segment, a ...
right-headed governing domain is established.
d. If V2 is a charmless segment which has a complexity greater
than VI' a right-headed governing domain is established.
e. If VI and V2 are charmed segments, no governing domain is
established.
58
.: .
CBAPTER~: OM GOVERNING RELATIONS BETWEEN NUCLEAR POSITIOHS
It follows from (6) that in a sequence of adjacent nuclear positions either

a left-headed or a right-headed governing domain is established; or no
governing relation is established. In the following pages we will consider
CHARETTE and YOSHIDA I S proposals on governing relations between adjacent
nuclear positions. They claim that a sequence of adjacent nuclear positions
is subject to the OCP.
5.2.1. Charette's proposal

According to the theory a sequence of adjacent nuclear positions is subject
to an interconstituent governing relation since the nuclear positions are
associated to distinct constituents (KLV (1988, 1990». Thus, a right-headed
governing domain is established. CHARETTE (1988) proposes that in a sequence
of adjacent nuclear positions the far left nuclear position is deleted by OCP.
Her proposal aims to account for the fact that in French the vowel of the
defini te article is deleted if the noun which follows it begins with a nuclear
position filled with a vowel, e.g. /la + ami/ surfaces as [lami] 'the friend'.
If the noun begins with a filled onset the vowel of the article remains, e.g.
/la + garse/ surfaces as [lagars5]. 'the boyl. In (7) we illustrate the forms
[lami] and [lagars5].
(7) a.
! I
I
R 0 R R 0
0 it 0
,I I I I
0
I I , R, I
"'II
,,, N,
I I I I I
I
I J( N , N I
I N
I I I I I I
I I I I I I I I I
X ~ X X X ---> X X X X
I
II I
I ,
I t
I
I
I
I
I
I
I
1 a a m 1 1 a m i
b.
0 R 0 R 0 R
,
I I
I
I
I 1\ I
I
I
I
I
I N I
I N\ I
I N
I
I ,
I I
I
I
I \ I
I
I
I
X X X X X X X
I
I
I
I
I
I
I
I
I
I ,
I
I a g a r s 5
In (7a) we have a sequence of strictly adjacent nuclear positions which is
subject to government. Charm requirements define that the noun-ini tial nuclear
position has the property of governing the nuclear position of the article.
A right-headed governing relation is established between the adjacent nuclear
positions and the deletion of the governed nuclear position, Le. "2he article-
final nuclear position, takes place as a consequence of the OCP.
2 Notice that in (7a) the governed nuclear position is deleted but the
segmental material which was formerly associated to it, i.e. a, remains as a
floating segment. The role of floating segments in phonological
59
CHAPTER 5: ON GOVERNING RELATIONS BETVEEN NUCLEAR POSITIONS
Vowel deletion does not take place in (7b) since no sequence of strictly
adjacent nuclear positions occurs so that no governing relation holds. In (8)
we summarize CHARETTE's proposal on the governing relation between nuclear
positions.
(8) Government Point deleted
by OCP
! I
I ! I
I
0 R 0 R 0 ~ 0 R
I I I
---- I ~--------- 1 I
N N If N
1
x x 1-
I
I
,
I 1
fi
I
I
vI V2 VI V2
In the following pages we will consider YOSHIDA's (1990) proposal which deals
with left-headed governing relations between adjacent nuclear positions.
5.2.2. Yoshida's proposal

, YOSHIDA (1990) proposes that a sequence of strictly adjacent nuclear positions
is subject to a process in which the two successive nuclear constituents are
fused into one by the application of the OCP. He calls this process "Nuclear
Fusion", which is illustrated in (9).
(9) N N N
I J.
I T I \
---) fi x ---) !. x
According to YOSHIDA (1990:344)
"When the two nuclei are merged into one, a branching nucleus is formed
and a constituent governing relation will be established between the
two nuclear points. In order for this process to be induced, there must
be no intervening non-nuclear segment between the two nuclei, and tne
same segmental requirements imposed on branching nuclei must be
satisfied. "
YOSHIDA's proposal illustrated in (9) aims to account for the fact that
Japanese forms which present a long vowel immediately followed by a geminated
consonant in careful speech; e.g. kootta, are frequently reduced. Either the
loss of the first half of the gemination, e.g. koota, or vowel shortening,
e.g. kotta, occurs. In (10) we illustrate YOSHIDA's analysis of the forms
kootta, koota and kotta.
representations will be discussed later in this Chapter.

60
.: .
eRAPTER 5: ON GOVERNING RELATIONS BETWEEN NUCLEAR POSITIONS
(10) a. 0 R 0 R 0 R
I
I
I
I 1\ I
I
I
I
I
I N N \ I N
I 1 I I
I I \ I 1
X X X X X X
I I
I \ / \1 I
k 0 t a
b. 0 R 0 R
1 I
I 1\ I
I I I
I I \ I
I
I N \ N
I
1 l\ \
x ! x (x) x X
I I
I \I \1 I
k 0 t a
c. - - 0 R 0 R
I
1 I
I
\ I
I ,
I
I I I I
I I \ I I
1
I N \ ·1 N
I t I
1 1\
1 . \ I I
X ! (x) ::It X X
I I
I \/' \1 I
k 0 t a
In -(lOa) there are two distinct nuclei linked to the segment Q. In careful
speech no change occurs and the form kootta - which presents a long vowel
followed by a geminated consonant - is derived.
(lOb) illustrates the representation of the form koota in which the long vowel
remains but the first half of the gemination is lost. According to YOSHIDA the
sequence of adjacent nuclear positions undergoes nuclear fusion and a
branching nucleus is formed so that the successive vowels 00 become a true
long vowel. Notice that in (lOb) the strict adjacency condition (which
determines that the head must be strictly adjacent to its complement (cf.
Chapter 1» is violated. This is because the underlined head of the rimal
consti tuent is not strictly adjacent to the rimal complement (which is
associated to the first half of the gemination). To avoid a violation of the
strict adjacency condition, YOSHIDA proposes that· the rimal pasi tion (in
parenthesis) is deleted. The form koota is derived.
In (IOc) we illustrate the representation of the form kotta which displays
vowel shortening while retaining gemination. YOSHIDA proposes that the nuclear
complement (in parenthesis) is deleted, so that the underlined head of the
rime can govern the rimal complement. Once the nuclear complement is deleted
the governing head and the rimal position are adjacent so that the strict
adjacency condition is respected.
In sum, YOSHIDA'S proposal on "Nuclear Fusion" (cf. (9» argues that in a
sequence of strictly adjacent nuclear positions where the far left one has the
property of governing the nuclear position to the right a constituent
governing relation is established and the two nuclear constituents are fused
61
.: .
CHAPTER 5: ON GOVERNING RELATIONS BETWEEN HUCLEAR POSITIONS
into a branching nucleus governing domain by OCP effects.

In the following section we will address some theoretical issues involving
governing relations between nuclear positions which arise from CHARETTE and
YOSHIDA's proposals. Then, in the final section of this Chapter we will extend
the previous proposals discussed earlier in this section and propose a general
analysis of governing relations between nuclear positions.
5.2.3. Conclusion
CHARETTE and YOSHIDA's proposals aimed to account for French and Japanese
phenomena respectively. However, nei ther of these works speci f i cally addresses
the theoretical issues which arise from their analysis. In the following pages
we will see that CHARETTE and YOSHIDA's proposals are in conflict with a basic
principle of GP, namely the projection principle. We will also see that their
works fail to define the conditions under which the OCP applies to a sequence
of adjacent nuclear positions.
We have seen that CHARETTE (1988) claims that in a right-headed nuclear
governing domain the governed nuclear position is deleted by OCP effects (cf.
Section 5.2.1). Nevertheless, to assume that the establishment of a governing
relation between nuclear position causes the deletion of the governed nuclear
posi tion viol ates one of the basic principles of the theory, namely the
projection princip~e.
(11) Projection Principle
Governing relations are defined at the level of lexical representation
and remain constant throughout a phonological derivation. (KLV
(1990:221»
According to the projection principle, a government relationship remain

constant throughout phonological derivations. That is, a governing relation
is an inalterable relationship. ~
CHARETTE's proposal that a governed nuclear position is deleted violates the

projection principle. -This is because if the governed nuclear position is
deleted, the previously defined governing relation no longer exists. Thus, a
governing relation is destroyed during the derivation violating the projection
principle.
YOSHIDA's proposal also encounters problems with the projection principle.
According to his proposal, either the complement of a branching nucleus or the
rimal complement is deleted during the process of deriving the reduced forms
koota and kotta (cf. (lOb,c)). Since a complement of a governing domain is
deleted during the derivational. process the projection principle is also
violated in his analysis.
We have just seen that CHARETTE and YOSHIDA's proposals are in conflict with
the projection principle. Nevertheless, their proposals explain French and
Japanese phenomena which involve sequences of adjacent nuclear positions.
Furthermore, their proposals also support the analysis of vowel-glide
alternations in BP which is presented later in this thesis:
62
CHAPTER 5: ON GOVERNING RELATIONS BETWEEN NUCLEAR POSITIONS
Whether it is the projection principle or CHARETTE and YOSHIDA's proposals

that need to be revised in order to avoid theory-internal conflict remains a
topic for future theoretical investigation. Throughout this thesis we will
adopt their proposals on governing relations between adjacent nuclear
positions with some amendments which are presented in the folloWing section.
These amendments concern the conditions under which the OCP applies.
We have seen that the OCP has been proposed to account for phonological
phenomena at the prosodic level (cf. section 5.1). However, CHARETTE and
YOSHIDA's proposals assume that the OCP applies at the skeletal and
... -cons.ti.tuent--l.e.vels-.- ..1Lthe--OCP_applies__ at__the...skelet.aLand .constauent_leveJ.s.. _
in a sequence of adjacent nuclear positions, we have to determine the
conditions under which it applies. This is because there are cases in which
a sequence of adjacent nuclear positions occurs in phonological
representations (cf. for example (lOa» and remain as such throughout the
derivation. Therefore, the OCP does not apply at random to all sequences of
adjacent nuclear positions. In the following pages we will expand CHARETTE and
YOSHIDA's proposals defining the conditions under which the OCP applies to a
sequence of adjacent nuclear positions.
5.3. Internuclear Goyernment

In this section we will present a general proposal for the analysis of
governing relations between adjacent nuclear positions. We will refer to these
governing relations as instances of internuclear government. We propose that
a
in sequence of strictly adjacent nuclear positions the OCP (Which prohibi ts
a sequence of identical phonological units from occurring) applies under
government conditions. We argue that once a governing domain is established
between a sequence of adjacent nuclear positions the OCP applies so that no
sequence of adjacent nuclear positions is present. If no governing relation
holds between a sequence of adjacent nuclear positions, the OCP (which we
claim applies under government) does not apply and a sequence of adjacent
nuclear positions is allowed to occur. An internuclear governing domain may
be either right or left-headed. In (12) we consider a right-headed
internuclear governing domain.
(12 )
a. Government b. Point deleted c. Light diphthong
under government formation
~ ~
I I
I I
0 R 0 R 0 Jt 0 R 0 R
I I I
I I 1 I
I I
N N )r N N
1 I
I
I
I
..
X ! ;. ~ x
I • I
I
I
I 1 I
I I
VI V2 VI V2 VI V2
63
.: .
CHAPTER S: ON GOVERNING RELATIONS BETWEEN NUCLEAR POSITIONS
For a right-headed internuclear governing relation to be established, the

segment associated to the governing position, Le. V2, must either be a
positively charmed segment or a complex charmless segment. The segment
associated to the governed position, Le. VI' must either be a complex
charmless segment 30r a simplex charmless segment or the governed nuclear
posi tion is empty. If VI and V2 are both complex charmless segments then V2 -
which is the governor - must have a complexity greater than VI - which is the
governee (cf. Chapter 1).
We propose that the OCP applies under government. Thus, the governed nuclear
posi tion is deleted by OCP effects (cf. (12b», so that no sequence of
identical phonological units, i.e. no sequence of adjacent nuclear positions,
-is present. --Notice--that--the--segment-a-l--materi-a-l·-whi-ch- was---formerly--as-s-oci-ated-·
to the governed nuclear position, i.e. VI' remains as a floating segment. We
propose that if the floating VI is a simplex charmless segment it relints to
the governing nuclear position forming a light diphthong (cf. (12c». If
the floating Vl is a complex charmless segment, it rema~ns a floating segment
which has no effect over phonological representations.
The fact that only simplex charmless segments - not complex charmless segments
- may relink to the governing nuclear position to form a light diphthong (cf.
(12c» follows from the segmental requirements imposed on light diphthongs.
Light diphthongs require the left member of the segmental sequence, i.e. Vj ,
to be a simplex charmless segment, i.e. [i,u,~]. The right member of the
segmental seque~ce in a light diphthong, i.e. v2, ray either be a complex
charmless segment or a positively charmed segment. Let us now consider a
left-headed internuclear governing domain. Consider (13).
3 For details of governing properties of nuclear segments cf. Chapter 1.

A summary of the governing properties of nuclear segments is presented in (5).
4 A light diphthong corresponds to what has been called in the literature
a rising diphthong. According to KAYE (1985:289) "a light diphthong is simply
a non-branching nucleus associated with two segments. For details on light
diphthongs within the GP framework see KAYE (1985, 1990b), KAYE & LOWENSTAMM
(1984) and CHARETTE (1988).
5 French illustrates instances of right-headed internuclear governing
relations where the floating segment is a complex charmless segment, i~e. ~,
which has no effect over phonological representations (cf. (7a».
6 For details on segmental requirements imposed on light diphthongs
within the GP framework see KAYE (1985,1990b) and KAYE & LOWENSTAMM (1981).
64
CHAPTER 5: ON GOVERNING RELATIONS BETWEEN KUCLEAR POSITIONS
(13)
a. Government b. Constituent c.Branching nucleus
dissociation under formation
government
,
I
! I
1 !
0 R 0 R 0 R 0 R 0 R
I 1 I I I
1 I I I I
N N N N N
I I I .1- I
I I I T I \
~ X ~ X ~ X
I I I I I I
I I 1 I I I
Vi V2 Vi V2 Vi V2
For a left-headed internuclear governing relation to be established, the
segment associated to the governing position, i . e. Vi' must either be a
charmed segment or a complex charmless segment. The governed position, i.e.
V2, must either be filled with a simplex charmless segment or must be an empty
nuclear position. We propose that the OCP applies under government. Thus, the
governed position dissociates from its nucleus and relinks to the governing
nucleus forming a branching nucleus (cf. (l3b». Noti ce that according to this
process no sequence of identical ph,ono1ogical units, i.e. no sequence of
adjacent nuclear positions, occurs. Indeed, we have two nuclear positions
bound together within a branching nucleus governing domain. When the governed
posi tion in a left-headed internuclear governing domain is empty the branching
nucleus corresponds to the representation of a long vowel where the governing
segment VI spreads over the governed position. When the governed position in
a left-headed internuclear governing domain is filled wi th a simplex charmless
segment thf branching nucleus corresponds to the representation of a heavy-
diphthong. Let us now consider the cases in which a sequence of adjacent
nuclear positions is not subject to government. Consider (14).
(14) 0 R 0 R
I I
I I
N N
I I
I I
X X
I
I ,
I
VI V2
In a sequence of adjacent nuclear positions such as in (14) where V1 and V2

are both charmed segments, no governing relation is established (because
charmed segments cannot be governed). We propose that in a sequence of
adjacent nuclear positions the OCP applies under government. Thus, i f no
governing relation holds between a sequence of adjacent nuclear positions the
OCP does not apply so that a sequence of adjacent nuclear posi tions is allowed
to occur.
7 Heavy-diphthongs represent what has been referred to in the literature

as falling diphthongs, i.e. vowel-glide sequences. A detailed discussion of
heavy-diphthongs in BP is provided in Chapter 7.
65
.: .
CHAPTER 5: ON GOVBRNING RELATIONS BETWEEN NUCLBAR POSITIONS
A sequence of adjacent nuclear positions is also allowed to occur if Vl

governs V2, i.e. a left-headed governing domain is defined, but V2 (which is
the governee) is a complex charmless segment. This is because the
establishment of a left-headed internuclear governing domain (which yields a
branching nucleus formation) requires that the governed nuclear position be
filled with a simplex charmless segment (cf. (13)). If the governed position
were filled with a complex charmless segment, charm requirements imposed on
branching nuclei would be violated, yielding an illicit representation.
Let us finally address the relationship between the metrical structure and the
establishment of internuclear governing relations. We argue that metrical
trees are built onto the projection of nuclear heads after the application of
phonological processes (cf. Chapter 2). We assume that only the nuclear heads
which are not governed are projected to the level of nuclear projection where
the metrical structure, i.e. secondary stress and word trees, is built.
Notice that the establishment of an internuclear governing relation triggers
either the deletion of the governed nuclear position (when a right-headed
internuclear governing domain is defined (cf. (12)) or the governed nuclear
position is incorporated into a branching nucleus structure (when a left-
headed internuclear governing domain is established (cf. (13)). Therefore,
given that a governed nuclear position is either deleted or incorporated into
a branching nucleus structure it will not be projected onto the level of
nuclear projection where the metrical structure is built. In Chapters 6 to 8 -
where we address the processes of vowel-glide alternations in BP - we will
illustrate how the metrical structure is built after phonological processes·
involving internuclear governing relations have taken place.
In this section we have set out the conditions under which an internuclear
governing relation mayor may not be established. We have proposed that a
right-headed internuclear governing relation yields a light diphthong
formation and that a left-headed internuclear governing relation yields a
branching nucleus formation. We have also seen that a sequence of adjacent
nuclear posi tiona is allowed to occur, 1. e. no internuclear governing relation
is established when either both nuclear positions are filled with a charmed
segment or the potential governed position in a left-headed internuclear
governing domain is filled with a complex charmless segment. We have also
argued that the metrical structure is built onto the level of nuclear
projection after the application of phonological processes.
5.4. Conclusion
In this Chapter we have considered governing relations between adjacent
nuclear positions. We have discussed earlier proposals on this topic and the
theoretical issues which arise from them. Then, we presented our proposal on
internuclear government which is an expansion of the earlier treatments of
governing relations between adjacent nuclear positions. We claimed that in a
sequence of adjacent nuclear positions the OCP applies under government. The
proposal presented in this Chapter will be applied to the analysis of vowel-
glide alternations in BP in the remaining Chapters of this thesis.
66
...: .
CHAPTER 6: PREVOCALIC GLIDES AND PREVOCALIC HIGH VOWELS
CHAPTER 6
PREVOCALIC GLIDES AND PREVOCALIC
HIGH VOWELS
6.0. Introduction
In this Chapter we will investigate forms in which prevocalic high vowels and
prevocalic glides occur in BP. We will refer to these forms as iV and uV
sequences, where y'stands for any vowel. We aim to define the conditions under
which a prevocalic high vowel alternates or fails to alternate with its
corresponding glide and we will also determine how glides are syllabified in
BP. In the first section we discuss the phonological representation of
prevocalic high vowels and prevocalic glides and 1n the following sections we
consider the behaviour of iV and uV sequences in relation to primary stress.
6.1. The Phonological Representation of Prevocalic Glides and Prevocalic High

Vowels
In this section we investigate the phonological representation of prevocalic
glides and prevocalic high vowels. According to the theory, neutrally charmed
elements such as 1° and UU (which correspond to the simplex charmless segments
i/ft) may occur either in a nuclear' head position or in a position which is not
a nuclear head. If i/ft occupy a nuclear head position a lax high vowel is
realized, i.e. [i,ft], whereas if i/ft occupy a position other than a nuclear
head a glide is manifested, i.e. [j,w]. Consider the representations in (1).
(1) a.O N b. 0 N c. 0 W
J
I
x x
/\ I
I
~ x X
J\ . I
I
I
I
I
I
t
I ,
I
i/\i V i/\i V C ~/O. V
d. 0 N e. 0 N 0 N
I I
I I
X X X .X
I I
/ \ II I I
C i/\i V i/\i V
Let us consider in detail each structure in (1). (la) represents a light

diphthong. According to KAYE (1985:2~9) "a light diphthong is simply a non-
branching nucleus associated with two segments". More precisely, it consists
of two segments associated to a single skeletal position. In metrical terms
a light diphthong behaves as a short vowel, Le.as a single nuclear
position. 1 Light diphthong require the left member of the segmental sequence
1 A light diphthong corresponds to what the phonological literature has

called a rising diphthong. We refer the reader to Chapter 7 where another type
of diphthong, namely the heavy diphthong (Le. falling diphthong), is
discussed. For a detailed discussion on light diphthongs see KAYE
67
to be a simplex charmless segment, Le. [i,O,~]. The right member of the

segmental sequence in a light diphthong may be either a complex charmless
segment or a positively charmed segment. Although two segments occupy a
nuclear position in the representation of a light diphthong, only the far
right member of the segmental sequence has the properties of a nuclear head.
Stress must be assigned to nuclear heads. Considering light diphthongs, it is
the far right segment which will bear stress. Therefore, in a light diphthong
the far right member of the segmental sequence is the'nuclear head, and the
far left one occurs in a position which is not a nuclear head.
(lb) illustrates a sequence of onset-nucleus in which the simplex charmless
segments i or ~ occupies the onset position. According to the theory, a non-
governing onset position may be filled either WiTh a negatively charmed
segment or with a charmless segment (cf. Chapter 1). Given that the segments
i/O are charmless they satisfy charm requirements for segments occurring in
onset positions.
(lc) illustrates a branching onset immediately followed by a nuclear position
which is filled with a vowel. The governing onset position must be filled with
a negatively charmed segment (which is represented by ~ in (lc». The governed
position in a branching onset must be filled with a charmless segment. Given
that the segments i/ii are charmless, they satisfy charm requirements for
segments'occurring in the governed position of a branching onset.
(ld), illustrates an onset position filled with a complex consonant immediately

followed by a nuclear position that is filled with a vowel. In the
representation of a complex consonant illustrated in (ld) the segments i or
g together with a consonantal segment are both associated to a single onset
posi tion. Complex consonants present restrictions wi ~h respect to which
segments may occur in the consonantal position, i.e. ~.
(Ie) illustrates a sequence of adjacent nuclear positions where the far left
nuclear position is filled with either the segment i or with the segment ~.
Notice that the two nuclear positions in (Ie) are adjacent to each other,
gi ven that the onset which intervenes between them does not dominate a
skeletal position.
Concerning the phonetic characteristic of the structures illustrated in (1)
it is important to mention that (la-d) have the same interpretation. That is,
in (la-d) the simplex charmless segments i and g -whose internal
representations consist of ·the elements 1° and UO respectively - are
pronounced as a glide, Le. [j,w]. On the other hand in (1e) the simplex
charmless segments i/fi are phonetically manifested as lax high vowels, i.e.
(1985,1990b).
2 A governing onset position, i.e. an onset position which is the head
in a branching onset governing domain, must be fi lIed with a negatively
charmed segment (cf. Chapter 1).
3 In section 6.3 we will discuss in detail the representation of complex
consonants.
68
·~ .
CHAPfER 6: PREVOCALIC ~LIDES AND PREVOCALIC BIGH VOWELS
From the lexical representation illustrated above we will determine the

conditions under which a high vowel or a glide is phonetically manifested. We
will argue that it is the governing relations es~ablished between strictly
adjacent nuclear positions that will determine wh~ther a high vowel or a glide
occurs.
Recall that government involves a relationship between two strictly adjacent
positions (cf. Chapter 1). In the representation illustrated in (6), we have
a sequence of strictly adjacent nuclear positions, given that the onset which
intervenes between them dominates no skeletal position. Therefore, they are
----------su-b-3-eGt---tQ--gQ\le-mmen-t-.-l-n-a----gG-v.e-m-i-ng~Gma-i_n_Gne_Qf-t_he_pos-i-t--i-Gns--i-s-the
governor - or the head - and the other position is the governee - or the
complement (cf. Chapter 1). The property of a skeletal position to govern
another skeletal position is determined by the charm value of the segments
which occupy the adjacent positions. That is, the governing position must be
filled with a segment which has the adequate charm value to govern the segment
associated to the governed posi tion. Notice that the far left nuclear posi tion
in (6) is filled with either of the simplex charmless segments Vii - whose
internal representation consists of the element 1° and UU respectively. We
have seen in Chapter 1 that the simplex charmless segments [i, ii] may be
governed by any other vowel of BP, i.e. [a,e,E,i,o,~,u,u,i,a]. Thus, in the
representation illustrated in (6) the far right nuclear position - which is
filled with any vowel of BP - is the governor and the-nuclear position to the
left - which is filled with either of the elements 1°/00 - is the complement.
A right-headed internuclear governing domain is established. _ -
In Chapter 5 we proposed that the establishment of a right-headed internuclear
governing relation results in the deletion of the governed nuclear position,
causing its segmental material to relink to the governing nuclear head
position forming a light diphthong. This process 1s illustrated in (7).
(7)
a. Government b. Point deleted c. Light diphthong
under government formation
~
I
I !
,
I
0 R 0 R 0 it 0 R 0 R
I I I I
I I 1 I' I
N N )f N N
I
I ,,; I
I
X ~ j.
I
I
I
I
~
I
I •
.' ,
X
I
1° /UO V IO/J V 1°/00 V
4"
According to our proposal illustrated in (7) the phonological
representation of glide-vowel sequences in BP corresponds to the structure of
a light diphthong. That is, either the element 1° or UO together with the
vowel are both associated to the same nuclear position. Given that in (7) the
elements 1°/00 occupy a position which is not a nuclear head, i.e. they occur
as the left-member of the segmental sequence in a light diphthong, they are
phonetically manifested as a glide, i.e. [j,w].
71
In the following pages we will provide evidence for this proposal, i.e. we
will show that glide-vowel sequences in BP correspond to the structure of a
light diphthong. We will then determine the conditions under which a glide or
a high vowel occurs. Let us first consider the phonological interpretation of
word-ini tial glide-vowel sequences, e.g. [jAtli] 'hiatus I and [werekema]
'Amazonian Indian group'. Evidence that word-initial glide-vowel sequences in
BP correspond to the structure of a light diphthong comes from a phonological
process which takes place in the Carioca dialect of BP (spoken in the state
of Rio de Janeiro). In this dialect s-final nouns are phonetical\y manifested
with a final palatal fricative, e.g. fmesf --> [meS] 'month'. When an s-
final noun is followed by a noun which begins with a filled onset, e.g. fmes
+ pasadli/ 'last month', the final § of the first noun appears as a palatal
'---rrlcaflve, l.e. [mespasadu]. In other worC1S; an s=final noun-ls reallzed as ,"--.-.------
a palatal fricative when it isSfollowed by a noun whose initial onset is
filled with segmental material. On the other hand, if an s-final noun is
followed by a noun which begins with a nuclear position, e.g. fmes + interlif
'whole month', the final ~ of the first noun is realized as an alveolar
fricative, i.e. [meziterli], but *[meSiterli] or *[me~iterli].
Let us consider the cases in which glide-initial nouns are preceded by an s-
final noun. Notice that if the glide occupies an onset position, we expect
that, as in forms which present an initial f~lled onset, the final § of the
first noun will be phonetically manifested as a palatal fricative (cf. fmes
+ pasadlif --) [meSpasadft]). On the other hand, if the glide occupies a nuclear
position we expect the final ~ of the initial noun to behave like a vowel-
initial noun, i.e. the final § of the first noun will appear as an alveolar
fricative (cf. fmes + interlif --> [meziterft]). Considering sequences of nouns
such as /tres + jat(isf 'three hiatus' or /tres + werekemas/ 'three Uerekemas'
- where an s-final noun is followed by a glide-vowel sequence - we observe
that the final § of the first noun is phonetically manifested as an alveolar
fricative, i.e. [trezjatliS] and [trezwerekemaS]. The fact that the final § of
the first noun is realized as an alveolar fricative when it is followed by a
glide-initial noun, i.e. [trezjatliS] and [trezwerekemaS], is ~vidence that the
phonological representation of word-initial glide-vowel sequences in BP
corresponds to the structure of a light diphthong. That is, both the glide and
the vowel are associated to a single nuclear position. This is because glide-
initial nouns behave like vowel-initial nouns when they are preceded by an s-
final noun. In both cases the final s of the first noun is phonetically
manifested as an alveolar fricative,- e.g. [meziterli] 'whole month' and
[trezjAtliS] 'three hiatus'; [trezwerekemaS] 'three Uerekemas'. In (8) we
illustrate the representation of the forms [jatli] 'hiatus' and {werekema]
'Uerekema I , where the glide-vowel sequence is syllabified iIi the
representation of a light diphthong.
4 In the Belo Horizonte dialect an alveolar fricative occurs word-

finally, e.g. fmesf --> [mes] 'month'.
S Indeed, an s-final noun occurs as a palatal fricative when it is
followed by a noun which begins with any possible word-initial consonant of
BP, i.e. [p,t,k,b,d,g,m,n,f,v,s,z,S,~,h,l].
72
. CHAPfER 6: PRRVOCALIC GLIDES AND PREVOCALIC HIGH VOVELS~
(8 )
a R a R a R a R a R a R
I I I I I I I I I I
N
I I
I
I
I
N
. I
N
I
I
I
I
N
,
I
I
I
N
I
I
I
I
N
I I I I I I I· I
I I I I I I I I
X
/ \ :
X X
I
I
X
/ \ I :
X X
I
I
X
I
I
X
I
I
X
,x
I
1° a t U UO e r' e k e m a
Let us now consider the phonological interpretation of glide-vowel sequences
.-----.....-.--p·r.eGeded-by-a-GQnsQnan-t-,-e-.-g-.-tp-i-ada-]-'3.Q·ke-'--[-kwalu-]--'GU-~d-'-.-I-n-(-9-)-we-···_-··--
illustrate the possible representations of these farms.
(9)
a.a R 0 R b. a R a R c. '0 .R o R
I
I , ,
I I I
I :\ I I II : I I I
I
I
N I
N : \ N : N I N
I I
I ~ I
I
I
I
I
1
,
I
I I \ I I : J .,, I
X X X X x x x x X )( X Jt X
I I
/ \ I
,
I
1
I
I
I
'0
I
I
I
I
I
I J /\ I I
p 1° a d a P I a d a p 1° a d a
kUO a A u k UO a A U k UO a A U
(9a) illustrates the case where the element 1° or UO together with a
consonantal segment occurs in the representation of a complex consonant. (9b)
illustrates the case where. the element 1° or UO occupies the governed posi tion
in a branching onset. (9c) illustrates the Case where the element 1° or UO
occurs in the representation of a light diphthong. In all representations in
(9) the elements 1° and UO are phonetically manifested as a glide since they
do not occur in a nuclear head position.
We have seen that word-initial glide-vowel sequences correspond to the
structure of a light diphthong (cf. (8». Similarly, we expect glide-vowel
sequences preceded by a consonant to correspond to the structure of a light
diphthong (cf. (9c». Thus, we have to show that representations like {9a-b)
do not correspond to the structure of glide-vowel sequences preceded by a
consonant in BP. If glide-vowel sequences correspond to the representation of
a complex. consonant - as illustrated in (9a) - we expect to find segmental
restrictions imposed on the consonant-glide sequence. This is because complex
consonants present restrictions with respect to the segmental materials, which
may occupy their skeletal position. However, this is not the case. That is,
no segmental restriction is imposed on consonant-gltde sequences where the
glide alternates with its corresponding high vowel. In fact any consonant
of BP may precede a glide-vowel sequence which alternates with a sequence of
~owels. Given that there is no segmental restriction as to which consonantal
material may precede the glide-vowel sequence (where the glide alternates wi th
6 In the. following section we will consider forms which present

consonant-gl ide sequences where the glide does not alternate with its
corresponding high vowel, e.g. [kwadru] but *[kuadru] 'picture'. We will see·
that in this case segmental restrictions are imposed on the consonant-glide
sequence.
73
,.
its corresponding high vowel) we exclude the structure of a complex consonant

- illustrated in (9a) - for forms which present glide-vowel sequences preceded
by a consonant.
Let 'us now consider the representation illustrated in (9b). If glide-vowel

sequences preceded by a consonant represent the element 1° or UO syllabified
in the governed position of a branching onset we do not expect to find forms
where the onset which precedes the glide-vowel sequence is branching, e.g.
[krjadu] -or [brwaka]. This is because constituents are maximally binary (cf.
Chapter 1). Thus, in a form where the glide-vowel sequence is preceded by a
branching onset, the glide cannot be syllabified in the onset position, which
is already branching, i.e. the onset is maximised. That is, a representation
-----TIke (10 J is impossible for forms such as [krjadii] and [brwaka].
(10) 0 R 0 R
* 1\,\ I
I i \ .
\
1,\
N I
W
; \'\
X X X
I
I
X
I
I
X
I
I
X
I I I I I
I I I I I
k r- I° a d li
b r UO a k a
Howeve~, forms presenting branching onsets followed by glide-vowel sequences
do occur in BP, as illustrated below~
(11 ) a. [krjadli] 'servant'
b. [frjeza) 'coldness'
c. [bibljotEka] , library'
d. [proprjedadi] 'property'
e. [brwaka] 'saddle-bag'
f. [flweti ] 'fluent I
g. [krwewdadi] 'cruelty'
h. [mestrwada] 'menstruating'
The fact that glide-vowel sequences may be preceded by branching onsets in BP
is evidence that the phonological representation of glide-vowel sequences
preceded by a consonant does not correspond to the structure of a branching
onset. This is because if glide-vowel sequences preceded by a consonant
corresponded to the structure of a branching onset, we would not find forms
where a branching onset precedes the glide-vowel seqUence. That is not the
case, as illustrated by forms in (11). The fact that glide-vowel sequences
preceded by a consonant do not correspond to the representation of a complex
consonant (cf. (9a», neither to the representation of a branching onset (cf.
(9b» is evidence that glide-vowel sequences preceded by a consonant
correspond to the structure of a light diphthong. That is, both the glide and
the vowel occupy a single nuclear position (cf. (9c». In (12) we illustrate
the representations of the forms [pjada] 'joke' and [brwaka) 'saddle-bag',
which present a glide-vowel sequence preceded by a consonant.
74
CHAPfER 6: PREVOCALIC GLIDES AND PREVOCALIC HIGH VOWELS
(12)
a. 0 R 0 R b. 0 R 0 R
I I
I
I
I
I
I
I 1\ I I
I
I
I N N I \ N N
I I I I I . I
I I I I \ I I
X X .X x x x x x X
I
I t\
I a d
I
I
I
a
I
I
I
I / \
I
I
r t;i': cJ. K
I
I
a-
P b
In the preceding pages we have seen that glide-vowel sequences in BP
_.. ---- -----GQ~lOes_penQ-te-t-he-
s-t-x:uG-tu-Fe-e-f-a--l-i-g-ht-d-i-ph-t-heng-.--ReGa-l-l-tha-t-any-f-Q-x:m--wh-i-Gh-------
presents pretonic glide-vowel sequences may also occur as a sequence of
vowels, i.e. [jV] - [iV] and [wV] - [uV] (cf. (4». We have argued that the
process of vowel-glide alternations in BP takes place as a consequence of the
establishment of a governing relation between adjacent nuclear positions (cf.
(7». In the following pages we will consider the derivation of the forms
[piada] - [pjada] 'joke'. We assume that any form which presents a pretonic
prevocalic glide that alternates with a pretonicprevocalic high vowel is
derived in a similar manner. Consider thf lexical representation of the forms
[piada] - [pjada] 'joke' given in (13).
(13 )
/ \
s w .'
0 R 0 R 0 R
I I I I I
I . I I I I
I I
I N N I N
I I I I I
I I I I I
X X X X X
I I I I I
I I I I I
P i a d a
1° At
From the lexical representation illustrated in (13) we will determine the

conditions under which a glide-vowel sequence, i.e. [pj&da], or a sequence of
vowels, Le. [pi&da], occurs. Let us first consider the case in which a glide-
vowel sequence occurs, i.e. [pjada]. We propose that glide-vowel sequences in
BP are deri.ved from the establishment of a right-headed internuclear governing
relation (cf. (7». The establishment of right-headed internuclear governing
relation results in the deletion of the governed .nuclear position. Then, .the
segmental material which formerly occupied the governed nuclear posi hon
relinks to the governing nuclear position forming a light diphthong. This
7 Recall that primary stress is lexically assigned (cf. Chapter 2).

Secondary and word level stress are assigned after phonological processes take
place. Therefore, the final metrical structure is built onto derived forms.
The phonological expressions which appear below the segments associated to the -
nuclear positions in (13) correspond to the internal representation of those
segments.
75
.; .
process is illustrated in (14).8

(14) a. b.
./\. .
__~S JfJ /"'\.
! :: s w
o ~O R 0 R 0 R 0 R
I I I I I I I I
I ~ N INN : N
I 1 : J I I I I
x t x x x x x x X
I
,
I I
I
I
I
I
I
I
I·
I
I
I
I
I
I
---- -------------------p--i----a;-----d---a--------p-i---a--a--a--------------------------------------- --- ------ --- ------ ------ ----
In (14b) the segment 1 occupies a position which is not a nuclear head, i.e.
it occurs as the left member of the segmental sequence in a light diphthong.
Given that the element 1° occupies a position which is not a nuclear head, it
appears as a palatal glide. The form [pjada] - which presents a glide-vowel
sequence - is derived and the metrical structure is built onto the projection
of nuclear heads. This is illustrat~d in (15).
Let us now consider the derivation of the form [piada], i.e. where a sequence
of vowels occurs. We reproduce its lexical representation in '(16).
8 In pretonic iV and uV sequences we will have a nuclear position filled

wi th either the simplex charmless segments i Iii immediately followed by a
nuclear position filled with either a complex charmless segment, or a
positively charmed segment. Given that complex charmless segments and
posi ti ve1y charmed segments have the property of governing a simplex charmless
segment, a right-headed internuclear governing domain is established.
76
.: .
Recall that a sequence of adjacent nuclear positions is subject to government

(cf. (14». In (16) we have a sequence of strictly adjacent nuclear positions
where the nuclear position which is filled with the segment [a], has the
property of governing the nuclear position to the left, which is filled with
the segment 1. A right-headed internuclear governing relation may potentially
be established. According to our proposal, the establishment of a right-headed
internuclear governing relation triggers the deletion of the governed nuclear
position (cf. (14a». Thus, the governed nuclear position in (16) should be
deleted. However, in a form such as [pia.da] there are three nuclear positions,
i.e. the same as in its lexical representation. That is, no nuclear position
is deleted. Thus, we have ,to explain why in the process of deriving the form
[piada] from the lexical representation illustrated in (16), no nuclear
position is deleted. We claim that in order to prevent the loss of a nuclear
posi ti.on (which would. take place under the establishment of a right-headed
internuclear governing relation (cf. (14» the ATR element is added to the
internal representation of the simplex charmless segment [i]. This in turn
yields t~e positively charmed segment [i] - whose internal representation is
("'+.1°)+. This process is illustrated in (17).
(17 )
a. 0 R 0
I
R
I
s
/
0
\
R
I
w
b. 0
I
R 0
I
s /'
R 0
I
w
R
I
I I I I I I I
I
N N N I N N N
I I I I I
I I I I I
X X X X X ---> )( K K x-· ><
I I I I I I I I
I I I I I I I I
P i a d a p i a d a
1° A+ 1° A+
I
1+
...
In order for a governing relation to hold, one of the positions must be filled
wi th a segment which has the property of governing the segment that occurs in
9 In the case the ATR element is added to the internal representation of

the simplex charmless segment [u] the positively charmed segment [u] - whose
internal representation is (...+.~)+ - is manifested.
77
: .
the other position. In the sequence of adjacent nuclear positions in (17b) the
far right nuclear position is filled with the positively charmed segment [a].
The nuclear position which iHllllT~iate1y follows raJ is filled with the
positively charmed segment [i). Recall that positively charmed segments
may govern a charmless segment, but not another positively charmed segment
(cf. Cha~ter 1). That is, positively charmed segments cannot be governed.
Thus, given that the adjacent nuclear positions in (l7b) are both filled with
positively charmed segments, a governing relation cannot hold between them.
Given that no governing relation holds between the adjacent nuclear positions
in (17b), no nuclear position is deleted and a sequence of strictly adjacent
nuclear positions is allowed to occur. The form [piada) is derived and the
metrical structure is built onto the projection of nuclear heads. This is
illlfst-rated- in (18).
(18) w s word tree level

~
I
1
w foot level
S
,
I
S
,
I
, I
I I I
I I I
0 R 0 R 0 R
I
I
I
I
I
I
, .
I
I
I
I I
I N N I N
I I I' I I
I I I I I
X X X X X
I
I
I
I ,
I I
I
P i a d a
1° A+
I
...'+
In this section we have considered the derivation of forms which present
pretonic prevocalic glides, e.g. [pjada], and pretonic prevocalic high vowels,
e.g. [pia-da]. Whether a prevocalic glide or a prevocalic high vowel is
phonetically manifested depends on whether or not a right-headed internuclear
governing relation is established. More specifically, we showed that pretonic
iV/uV sequences are subject to the following processes:
a) A right-headed internuclear governing domain is established
resul ting in the deletion of the governed nuclear position. The
segmental material which was formerly associated to the governed
position, i.e. the high vowel, relinks to the governing position
forming a light diphthong. A glide-vowel sequence is realized.
b) In order to prevent the loss of a nuclear position (which we
claim takes place under the establishment of a right-headed
10 Recall that pretonic nuclear heads must be filled with positively

charmed segments (cf. Chapter 3). In (17b) the nuclear position filled with
the tense high front vowel [i) - which is a positively charmed segment -
occurs in pretonic position so that charm requirements are satis.f~ed.
78
CHAPtER 6: PREVOCALIC GLIDES ANDoPRBVOCALIC HIGH VOWELS
internuclear governing domain) the ATR element 1s incorporated into the

internal representation of the high vowel. A tense high vowel - which
is a positively charmed segment - is realized. No governing relation
holds between the adjacent nuclear positions, and no nuclear position
. is deleted. A sequence of. adjacent nuclear positions is allowed Jo
occur, and a sequence of vowels is realized.
In this section we have seen that pretonic iV/uV sequences may appear either
as a sequence of vowels, i.e. [iV,uV], or as a glide-vowel sequence, i.e.
[jV,wV]. If all forms which present pretonic glide-vowel sequences behave in
_. e.xas:J:l.y: the same way, we can expect any form which presents a pretonic gl1de-
vowel sequence to alternate with a form in which a sequence of vowels occurs.
There are however some forms where a pretonic glide-vowel sequence occurs and
the glide does not alternate with its corresponding high vowel. A sample of
these forms is given in (19).
(19)
a. [kwadIii] * [kuadIii] 'picture'
b. [akwarEla] * [akuarEla] 'water colour'
c. [gwahda] * [guahda] 'guard'
d. [gwahdanapu] *[guahdanapu] ,serviette' .
In the following section we will consider forms such as the ones illustrated
in (19). We will show that in the cases where a pretonic glide does not
alternate with its corresponding high vowel, i.e. [kwadru] but *[kuadru], the
glide-vowel sequence is not derived from a sequence of adjacent nuclear
positions; but rather the glide presents a special type of syllabification.
6.3. The Syllabification of the Glide [wl

In this section we will consider forms where the glide ~ corresponds,to the
element uO leXically associated to a position other than a nuclear head.
Consider the possible syllabifications of the glide ~ in (20).
(20) a. 0 R b. 0 R c. 0 R
I I I
I 1\ I I
I
N I \ N N
I I I I
I I \ I . I
X X X X X x
I I I
/ \ II
J\ C
I
1f
I
V C uO V
The representations in (20) illustrate the element uO syllabified in a
position which is not a nuclear head. Thus, it will be phonetically manifested
as a glide, i.e. [yt. In (20a) the element uO occurs in the representation of
a light diphthong. In (20b) the element uO occurs in the governed position
11 Recall that when the element UO occurs as the left member of the
segmental sequence in a light diphthong it does not have the properties of a
nuclear head, e.g. it does not bear stress. Thus, although in the
representation of a light diphthong the element UO occurs in a nuclear
79
of a branching onset. In (20c) the element trO together with a consonant is

associated to a single onset position forming a complex segment.
In the following pages we will investigate which of the structures in (20)
correspond to the syllabification of the glide [w] in forms such as [kwadru]
'picture' (cf. (19». Let us first consider (20a) where the glide-vowel
sequence corresponds to the structure of a light diphthong. If in a form like
[kwadru] the glide-vowel sequence corresponds to the· structure of a light
diphthong, we have a lexical representation as in (21).
(21) / \
s w
-0 ---R---0----R---------····-----------·-·----------·-----.--_...- --.-----.-
I 1\ I
N I \ N
I i \
x x x x X
II / \ :,I1 " II
k trO a d r u
We claim that (21) does not correspond to the lexical representation of
[kwadru]. The first argument against a structure like (21) comes from the lack
. of restrictions concerning the vowel in the glide-vowel sequence. This is
because in languages which present lexical I ight diphthongs there are
constraints with respect to the vowel in glide-vowel seqUences which
correspond to light diphthongs (cf. KAYE (198ge».
Let us cOrfider French which is a language that has lexical light
diphthongs. Interestingly, in the glide-vowel sequences that correspond to
a light diphthong in French there are restrictions with respect to which
vowels can occur in the glide-vowel sequence. That is, the vowel can O~1Y be
(a,e,o] as in [wat] 'wad of cotton', [lw~] 'far' and (lwo] 'we rent'. If
in forms like [kwadru] the glide-vowel sequence corresponded to the structure
of a light dipbthong - as proposed in (21) - then there woul~4be constraints
concerning which vowels can follow the glide. Consider (22).
position it does not occupy a position which is a nuclear head (cf. Section
6.1) .
12 Evidence for lexical light diphthongs in French i~ given in KAYE &

LOWENSTAMM (1984); CHARETTE (1988) and KAYE (1989b).
13 Light diphthongs in French which are manifested as palatal glide-vowel
sequences may present other vowels than a,e,o in the vowel position, e.g. ~
[j~d] iodine' .
I
14 In all forms illustrated in (22) the glide-vowel sequence does not

alternate with a sequence of vowels, i.e. [wV] does not alternate with [uV].
Therefore, forms in (22) illustrate cases where the glide (w] corresponds to
the element trO leXically associated to a position other than a nuclear head.
80
CHAPTER 6: PREVOCALIC GLIDES AND PREVOCALIC BIGH VOWELS
(22) a. [gwahda] 'guard'

b. [kngweta] 'stool-pigeon'
c. [sekwEla] 'sequel'
d. [lrgwisa] 'sausage'
e. [akw6zii] 'aqueous'
f. [kw5ta] _'quota'
Forms in (22) show that a back glide can be followed by any BP vowel. 15 We
assume that the lack of constraints concerning the vowel in the glide-vowel
sequence in (22) - namely that any vowel can follow the glide - provides us
wi th evidence that the glide-vowel sequence does not correspond to the
structure of a light diphthong. This is because if the glide-vowel sequence
corresponded to the structure of a light diphthong, there would be constraints
with respect to which vowels could follow the glide, which is not the case.
More evidence that glide-vowel sequences do not correspond to the structure
of a light diphthong comes from the fact that glide-vowfll sequences in BP
cannot be preceded by branching onsets, e.g. *[krwadrfi]. Recall that in a
light diphthong both the glide and the vowel are associated to the same
nuclear position. Thus, given that the glide occurs in a nuclear position
there should be no restri~tion on whether the preceding onset is branching or
not. That is, we should find forms where the glide-vowel sequence1fs preceded
by a branching onset, e.g. *[krwadrii] - which is not the case. The facts
we have just discussed show that lexical glide-vowel sequences in BP do not
correspond to the structure of a light diphthong. That is, a representation
as the one illustrated in (20a) is excluded.
15 In (22) the vowel which follows the glide bears primary stress. Recall
that in primary stressed positions the following vowels occur in BP:
[a,E,e,i,o,~,u] (cf. Chapter 3). The only constraint in forms in (22) is that
the glide cannot be followed by the vowel [u], i.e. *[wu].
16 Recall that glide-vowel sequences which alternate with a sequence of
vowels may be-p-receded by- branching- ensets ie. g; [brwaka] - [bruaka-]' saddle-
bag' (cf. Section 6.2). As we have seen, in these forms the glide is derived
from a sequence of nuclei. In this section we are analysing forms where the
glide [w] corresponds to the element ua lexically associated to a position
other than a nuclear head. In these forms the glide-vowel sequence does not
alternate with a sequence of vowels.
17 If we consider French which is a language that has lexical light

diphthongs, we notice that a glide-vowel sequence which corresponds to the
structure of a light diphthong can be preceded by a branching onset, e.g.
[trwa] 'three'. The representation of [trwa] is illustrated below.
o R
1\ I
I \ N
: \ :
x x x
I : / \
t r ua a
81
Let us then consider (20b) which presents the element UO syllabified in the
governed position of a branching onset. If in a form like [kwadru] the glide
is syllabified in the governed position of a branching onset, we have a
representation such as the one illustrated below.
(23) / \
s w
0 R 0 R
I I
1\ I
,!\ I
I
I \ N I \ N
I
I \ ,
1 I
I \ I
I
X X X X X X
I I I I I I
I I I I I I
k UO a d r 6.
Assuming that the glide ~ represents the element UO syllabified in the
governed posit ion of a branching onset - as illustrated in (23) - we can
expect that whenever the glide ~ occurs it will be preceded by a consonant
(which will be the head of the branching onset governing domain). Indeed, a
consonant always precedes the glide in all forms where the glide ~corresponds
to the element UO lexically associated to a position other than a nuclear head
(cf. (19,22)). That is, a consonant-glide sequence, i.e. Cw, occurs. Recall
that constituents are maximally binary (cf. Chapter 1). That is, branching
constituents are at the most associated to two ske~etal positions. Thus, if
in a form like [kwadru] the glide represents the element tf' syllabified in the
governed position of a branching onset, the presence of the glide maximises
the onset. Given that constituents are maximally binary, we expect not to find
any forms where the glide ~ is preceded by a b.ranching onset. Tha\aiS, a form
like *[krwadru] would not occur - and that is exactly the case. Thus, if
we assume that the glide ~ represents the element UO syllabified in the
governed position of a branching onset (cf. (23)), we account for the lack of
forms where a branching onset precedes the glide ~, i.e. *[krwadr6.]. If we
a~sume that the glide ~ corresponds to the element UO syllabified in the
governed position of a branching onset, we expect forms which present a
consonant-glide sequence to behave like any branching onset in BP. Lei us
consider a phonological process which involves branching onsets in BP. 1 In
(24) we illustrate forms which present branching onsets,-Le. obstruent-liquid
sequences, followed by a primary stressed vowel.
18 Once again we stress that the cases under discussion in this section
involve those glide-vowel sequences which do not alternate with a48equence of
vowels (cf. (19,22)). According to our proposal in these forms the glide ~
corresponds to the element UO lexically associated to a position other than
a nuclear head. Recall that derived glide-vowel sequences may be preceded by
branching onsets, e.g. [brwaka] - [bruaka] !saddle-bag'.
19 For more details of this process see SILVA (1989b).
82
CHAPTER 6: PREVOCALIC GLIDES AND PREVOCALIC BIGB VOWELS
(24) a. [pratii] 'plate'

b. [frevii] 'frevo (dance)'
c. [brEvi] 'brief'
d. [krimi] 'crime'
e. [gr6sii] 'thick'
f. [tr:Ska] 'change'
g. [bluza] 'blouse'
Forms in (24) show that obstruent-liquid sequences can be follo~ed by any BP
primary stressed vowel. In (25) we illustrate forms where the obsfruent-liquid
sequence occurs -------
-- stress --------
~
followed by a nuclear
--.-.-.-.-. posi tionwhich
.- .--..- ----- does---.--
-.- - ..-.- ---.----- not - bear primary
- . -.- _.-. --.-.. --- --.
(25) a. [ez~plii] 'example'

b. [6tru] 'other'
c. [sepri] 'always'
d. [livrii] 'book'
e. [flam~gU] 'Flamengo'
f. [brazilerii] 'Brazilian'
g. [koprimidii] 'tablet'
h. [k5plikadii] 'complicated'
Forms in (2Sa-d) illustrate cases in which the obstruent-liquid sequence
occurs in a position that follows the primary stressed vowel. Forms in (25e-h)
illustrate cases in which the obstruent-liquid sequence occurs in a position
that precedes the primary stressed vowel. In all forms illustrated in (25),
i.e. where an obstruent-liquid sequence is followed by a vowel that does not
bear primary stress, the obstruent-liquid sequence may occur as a single
consonant~ i.e. only the onset head is realized. This process is illustrated
in (26).
(26) a. [ez~plft] [ezepii]
b. [livrii] [ livii]
. c. [brazilerii] [bazilerii]
d. [k5plikadii] [k5pikadft]
Forms in the left column in (26) illustrate the cases where an obstruent-
liquid sequence is pronounced. Forms in the far right column illustrate the
cases where only the obstruent· is phonetically manifested. The process
illustrated in (26) (where an obstruent-liquid sequence alternates with an
obstruent) takes place only if the vowel which follows the obstruent-liquid
sequence does not bear primary stress. In cases where obstruent-liquid
sequences are followed by a primary stressed vowel, e.g. [pratft] 'plate' (cf.
(24», an obstruent-liquid sequence must be phonetically manifested, i.e.
*[patu]. As we have just seen, a branching onset, i.e. an obstruent-liquid
sequence, alternates with a non-branching onset, Le. an obstruent. This
happens if ·the vowel which follows the branching onset does not bear primary
stress. Consider forms in (27), which present consonant-glide sequences.
20 The examples in (26) represent a sample of forms given in (25). All

forms in (25) may undergo the process illustrated in (26).
83
(27) a.[gwahdanapa] * [gahdanapo.] 'serviette'

b. [akwarEla] *[akarEla] 'water colour'
c. [ligwa] *I liga] 'language'
d. [inikwa] *[inika] 'iniquitous'
Forms in (27) illustrate consonant-glide sequences followed by a nuclear
position which does not bear primary stress. In (27a-b) the consonant-glide
sequence is followed by a nuclear position that precedes primary stress, and
in (27c-d) the consonant-glide sequence is followed by a nuclear position that
follows primary stress. If the consonant-glide sequences in (27) corresponded
to the structure of a branching onset - as illustrated in (23) - we would
expect consonant-glide sequences to alternate with a consonant when they are
----foU-owed---by -a--vowe-l--that--dcras--not -bear--pTtmary -stre-ss-;--However; thi-s--ts--not--
the case, as illustrated in forms in the far right column in (27). Thus, the
fact that the consonant-glide sequences in (27) do not alternate with a
consonant is evidence that the phonological representation of consonant-glide
sequences do not correspond to the structure of a branching onset.
Asecond argument against the branching onset hypothesis is that in consonant-
glide sequences the are restrictions with respect to which consonant can
precede the glide. 1f If we consider any form where the glide in the
consonant-glide sequence corresponds to the element UO lexically associated
to a position other than a nuclear head (cf. (19,22,27», we notice that in
all these forms the consonant which precedes the glide is a velar stop, i.e.
either [k] or [g]. The constraint with respect to the consonant in the
consonant-glide sequence (only a velar stop can occupy the consonantal
position) provides us with further evidence that the consonant-glide sequence
does not correspond to the structure of a branching onset- as illustrated in .
(23). If the consonant-glide sequences corresponded to the structure of a
branching onset, we would expect other consonants - besides the velar stops
[k, g] - to precede the glide. This is because we find obstruent-liquid
sequences - which correspond to branching onsets - presenting other consonants
besides velar stops, e.g. [plana] 'plan', [trigti] 'wheat'. Therefore, we
expect that if consonant-glide sequences corresponded to the structure of
branching onsets, the consonant-glide sequence would present other consonants
- besides the velar stops [k,g] - in the consonantal position. That is, *[pw]
and *[tw] would be possible sequences. This is not the case.
The facts we have just discussed above show that a consonant-glide sequence -
where the glide corresponds to the element UO lexically associate~ to a
position other than a nuclear head - does not correspond to the structure of .
a branching onset. We propose that the syllabification of consonant-glide -
sequences in BP - which represent a labio-velar consonant - correspond to a
complex segment as illustrated in (28).
21 Recall that in this section we are analysing consonan~glide sequences

where the glide corresponds to the element UO lexically associated to a
position other than a nuclear head. In consonant-glide sequences where the
glide is derived from a sequence of nuclei so that either a glide-vowel
sequence or a sequence of vowels occurs, e.g. [kwalu] - [kuala] 'curd', there
is no constraint with respect to the consonant which precedes the glide. That
is, any BP consonant may precede the glide.
84
.: .
(28) 0 R
I I
I I
I
I N
I I
I I
X x
I
/ \ I
C UO V
A complex segment presents two segments associated to a single onset position.

Thus, in the representation of consonant-glide sequences both the velar stop
and the glide ~ (whose internal represe~tation consists of the element 00) are
associated to the same onset position.
In this section we have considered the syllabification of the glide ~ when it
represents the element UO lexically associated to a position other than a
nuclear head. In the following section we return to the discussion of iV/uV
sequences. More specifically we will consider forms in which iV/uV sequences
occur posttonically.
6.4. Posttonic Prevocalic Glides and Postton1c POltvocalic High Vowels

In this section we consider forms in which iV/uV sequences occur in the final
nuclear positions in antepenultimate stressed nouns. That is, the weak
positions in the antepenultimate stress pattern sww. Consider (29).
(29) a. [sabja] 'wise'
b. [familja] 'family'
c. [Sljii] 'oil '
d. [edifisjii] 'building'
e. [agwa]~ 'water'
f. [ig~nwa] 'ingenuous'
g. [tenwi] 'tenuous'
h. [kogrwi] 'congruous'
Forms in (29) illustrate glide-vowel sequences occurring in a position which
follows primary stress. Recall that in posttonic posi tion the following vowels
occur: [a,i,u] (cf. Chapter 3). Therefore, we expect any of the posttonic
vowels [a,i,ii] to occupy the vowel position in posttonic iV/uV sequences. In
the iV sequences illustrated in (29a-d) the vowel position is filled with
either of the segments [a,ii] and in the uV sequences illustrate~ in (2ge-h)
the vqwel position is filled with either of the segments [a,i].
22. For theoretical issues concerning the representation of complex

consonants and some constraints imposed on complex consonants in BP see SILVA
(1992).
23 In Chapter 8 (Where governing relations between high vowels are
addressed) we will consider in detail forms presenting a glide followed by a
high vowel (cf. (29c-d); (29g-h)), and forms presenting posttonic iV/uV
sequences where a sequence of identical high vowels occurs.
85
CHAPTER 6: PRETONIC GLIDES AND PRETONIC HIGH VOWELS
It is important to mention that unlike pretonic iV/uV sequences - which may

occur either as a glide-vowel sequence or as a sequence of vowels, e.g.
[pjada] - [piAda] 'joke' and [kwa,Hi] - [kua),ii] 'curd' (cf. Section 6.2) -
posttonic i VluV sequences do not alternate with their corresponding high
vowel. That is, in posttonic iV/uV sequences (which are exemplified by the
forms in (29» a glide-vowel sequence must be realized, e.g. [sabja] but
*[sabia] and [agwa] but *[agiia]. As we will see in the following pages, such
a result is predicted by our analysis based on GP.
In the following pages we will provide evidence that in a similar way to
pretonic iV/uV sequences (which are manifested as glide-vowel sequences, cf.
section 6.2.) a posttonic iV/uV sequence (which occurs as a glide-vowi1
sequence) is derived from a sequence of adjacent nuclear positions.
Evidence that posttonic iV/uV sequences (which are realized as glide-vowel
sequences) are derived from a sequence of adjacent nuclear positions comes
from the behaviour of glide-vowel sequences with respect to the primary
stressed vowel. Primary stress in BP may be final, penultimate or
antepenul timate (cf. Chapter 2). Interestingly, in forms which present
.--- posttonic glide-vowel sequences, primary stress must fall on the nuclear
position which immediately precedes the glide-vowel sequence, i.e. [familja]
but *[familja] and [igenwa] but *[igenwa]. The fact that primary stress must
fall on the nuclear position which immediately precedes the glide-vowel
sequence is evidence that posttonic glide-vowel sequences are derived from a
sequence of nuclear positions. This is 'because the nuclear position which
immediately precedes the glide-vowel sequence corresponds to the third-to-last
nuclear position, fitting into the antepenultimate stress pattern.
Let us then consider how posttonic glide~vowel sequences are derived. 25 The
lexical representation of the form [sabja] 'wise l is illustrated in (30).
(30) ~
w
s w
0 R 0 R 0 R
I I I
I I I
N N N
r
I
X X X X X
I I I I I
I I I I I
S a b i a
rO vO
"1.\
AI
11"
24 There are cases in which the glide in posttonic glide-vowel sequences
is syllabified in the representation of a complex consonant. In these cases
a velar stop-glide sequence must occur, e.g. [ligwa] 'language'. These forms
are discussed later in this section.
25 We assume that any form in (29) - presenting posttonic iV/uV sequences
which occur as glide-vowel sequences - is analysed in like manner.
86
CHAPTER 6: PRETONIC GLIDES AND PRKTONIC HIGH VOWELS
Following our proposal we assume that posttonic glide-vowel sequences are

derived from the establishment of a right-headed internuclear governing
relation (cf. Chapter 5). The establishment of a right-headed internuclear
governing relation triggers the deletion of the governed nuclear position,
leaving its segmental material to relink to the governing nuclfiar position
forming a light diphthong. This process is illustrated in (31).
(31)
a. : f==::::::::--- b.
s w w
I
I
-_._--~_._. __ __ __._-----_._--------
. ..
I
J ! I
I s w
o R 0 ~ 0 R o ..~
I I
C
I
·R
I
I I I
I J 1 I I I I I
N I
I Jf N I N I N
I
I
I
I 1 I
I
I
I
I
I
I
I
X X x t X ---> )q.x.X.K
I I I I
I I I 1 I I : I / \
S a b i a s a b i a
IO
In (31) a right-headed internuclear governing relation is established and a

light diphthong is formed. The form [sabja] 'wise' - which presents a
posttonic glide-vowel sequence - is derived.
In the beginning of this section we pointed out that posttonic i V/uV sequences
must be phonetically manifested as a glide-vowel sequence, i.e. [sabja] but
*[sabia] and [agwa] but *[agfia]. The fact that posttonic iV/uV sequences must
be phonetically~manifested as a glide-vowel sequence is indeed predicted by
our analysis. Recall that posttonic nuclear positions must be filled with
charmless segments (cf. Chapter 3). Thus, in posttonic iV/uV sequences the
"nuclear position which immediately precedes the vowel must be filled with the
charmless segment i or £. On the other hand, the final nuclear position in
posttonic iV/uV sequences - which corresponds to the vowel position - may be
filled with any of the segments [a,u,i]. The simplex charmless segment i ~~
£ may be governed by any other nuclear segment of BP (cf. Chapter I).
Therefore, we predict that in posttonic iV/uV sequences a governing relation
will always hold between the nuclear position filled with the vowel and the
26 Recall that in a governing domain one of the posi tions must have
adequate charm value to govern the segment occurring in the other position. ~
The final nuclear position in posttonic iV/uV sequences (which may be filled
with either [a,u,i]) has the property of governing the immediately preceding
nuclear posi tion, which is filled wi th ei ther of the simplex charmless segment
i/u. For details on governing properties of nuclear segments see Chapter 1.
21 The fact that the simplex charmless segments, i.e. i/u, may govern
each other is addressed in Chapter 8, where governing relations between high
vowels are discussed.
87
CHAPTER 6: PRETOHIC GLIDES AND PRETONIC BIGB VOWELS
preceding nuclear position - which is filled with the simplex charmless

segment i or y. A right-headed internuclear governing domain is defined
yielding a light diphthong formation (cf. (31», so that a glide-vowel
sequence is manifested.
The fact that in posttonic iV/uV sequences the nuclear position associated to
the high vowel will be filled wi th one of the simplex charmless segments i/u -
which may be governed by any other nuclear segment of BP - explains why a
glide-vowel sequence must occur posttonically. That is, the establishment of
a right-headed internuclear governing relation in posttonic iV/uV sequences
must take place so that a glide-vowel sequence will occur. In the following
pages we will consider the phonological interpretation of forms which present
- . --pes-t-teR-i-e-g-l-ide-·vewel--sequenees--p;r;eseded-by··a-vela-~-s-tep-G(msQnan-t-.-GQn-s-i-de·r-
(32).
(32) a. [inikwa] 'iniquitous'
b. [ Irgwa] 'language'
c. [ inSkwa] 'innocuous'
d. [agwa] •water ,
Forms in (32) present posttonic glide-vowel sequences. Notice, however, that
in all these forms the glide-vowel sequence is preceded by a velar stop
consonant so that a velar stop-glide sequence occurs. We have seen that a
velar stop-glide sequence may be interpreted as a complex consonant, i.e. both
the velar stop and the element uO are syllabified in the same onset position
(cf. Section 6.3). Therefore, one could question whether the glide in the
glide-vowel sequences illustrated in (32) corresponds to the element uO
syllabified in the representation ofa complex consonant, or whether. the glide
represents the element uO syllabified in the representation of a light
diphthong (as a consequence of the establishment of a right-headed
internuclear governing relation). These two possibilities are illustrated in
(33) .
(33) a. 0 R b. 0 R
I I
I I
N N
x x x x
/ \ :\ I
I / \
C uO V C uOv
In order to determine how the glide ~ is interpreted in posttonic velar stop-
glide-vowel sequences, we have to consider derived forms where the velar stop-
glide-vowel sequence occurs in pretonic position. Recall that in cases where
a pretonic glide is derived from a sequence of adjacent nuclear positions it
alternates with its corresponding high vowel, e.g. [kwalu] - [kualu] 'curd'
(ct. section 6.2). On the other hand it the pret~ic glide corresponds to the
element UO syllabified in the representation of a complex consonant, it cannot
alternate with its corresponding high vowel, e.g. [kwadru] but * [kuadru]
'picture' (cf. section 6.3). Consider forms in (34).
88
CHAPTER 6: PRETONIC GLIDES AND PRETONIC BIGH VOWELS
(34)
a. [inikwi dlldi] *[inikuidlldi] 'iniquity'
b. [ligwistlka] *[liguistika] 'linguistics'
c.[inokwidlldi] - [inokuidadi] 'innocuousness'
d.[agwadu] - [aguadu] 'watered'
(34) illustrates derived forms of the nouns listed in (32). Forms in (34a-b)
show that the pretonic glide cannot alternate with its corresponding high
vowel. Forms in (34c-d) show that either a pretonic high vowel or a pretonic
glide may occur.
--,"---- __ • _ _
---,-"~._---_._._. - . _ . _•• ,_ •• __ • .,~ •• , _ ••• " _ _ ., • __ , _ , ••• _ _ ' . _ _ ,•• , • ._" __ _ _ _•• ' __ M _ . . . . _ . . ., _ . _ . • _._ ,~_ ", , __" '." .,, _ _ ._._ _ " • • ._. _ ••• ~ __ " .•• _
The derived forms illustrated in (34) lead us to determine the correct

phonological interpretation of the glide in forms which present posttonic
velar stop-glide-vowel sequences - as illustrated in (32a-b). That is, .in
forms such as (32a-b), i.e. [inikwa] and [IIgwa], the glide corresponds to the
element ~ syllabified in the representation of a complex consonant. This is
why in the derived forms [inikwidadi] and (ligwistika] (cf. (34a-b» the
pretonic glide cannot alternate with its corresponding high vowel; 1. e.
*(inikuidadi] and *[liguistika]. On the other hand, in forms such as (32c-d),
i.e. [in5kwa] and [llgwa], the glide corresponds to the element uO syllabified
in the representation of a light diphthong (as a consequence of the
establishment of a right-headed internuclear governing relatipn). This is why
in the derived forms [inokwidlldi] - [inokuidadi] and [agwadfi} - [aguadfi] the
pretonic glide alternates with its corresponding high vowel (cf. (34c-d».
In this section we have considered forms which present post tonic i V/uV
sequences. We have seen that posttonic iV/uV sequences must be phonetically
manifested as a glide-vowel sequence. Such a requirement, i.e. that a glide-
vowel sequence must occur posttonically, follows from charm requirements with
respect to the position of primary stress and the establishment of a right-
headed int~rnuclear governing relation. We have seen that the glide in
posttonic iV/uV sequences may either be derived from a sequence of adjacent
nuclear positions or it may correspon~ to the element UO syllabified in the
representation of a complex consonant. 8 In the following section we consider
28 Our argument that posttonic uV sequences are derived from a sequence

of nuclei is based on the behaviour of glide-vowel sequences with respect to
the primary stressed vowel. That is, primary stress must fallon the nuclear
position which immediately precedes the glide-vowel sequence, i.e. [tgenwa)
'ingenuous' but *[igenwa]. In (32) we illustrated forms which present
posttonic glide-vowel sequences. According to our proposal in (32a-b) the
glide corresponds to the element uP lexically associated to a position other
than a nuclear head and in (32c-d) the glide is derived from a sequence of
nuclear positions. Notice that we expect to find forms in which the gl~e
corresponds to the element uP lexically associated to a position other than
a nuclear head where primary stress falls on the second-to-last nuclear
position which precedes the glide. That is, [inikwa] should be a possible
form. This is because if the glide is not derived from a sequence of nuclei -
but it rather represents the element UO together with the velar stop
consonant associated to an onset position - nothing should prevent stress from
falling on the second-to-last nuclear position which precedes the glide.
89
CHAPTER 6: PRETONIC GLIDES AND PRETOHIC HIGH VOWELS
iV/uV sequences where the nuclear position filled with the high vowel bears
primary stress.
6.5. Prevoca11c Primary Stressed High Vowels

In the previous sections we analysed iV/uV sequences where a glide alternates
with a high vowel, i.e. in pretonic position (cf. Section 6.2); and iV/uV
sequences where a glide must occur, i.e. in posttonic position (cf. Section
6.4). In this section we will see that in primary stressed iV/uV sequences
ei ther a sequence of vowels or a vowel-glide sequence may be realized.
_. Conside~ (35J..!. __
(35) a. [pia] 'sink'
b. [ekonomia] 'economy'
c. [hua] 'street'
d. [kakatua] 'cacatua (bird)' . .... ...,.....
e. [maniaka] 'maniac'
f. [oUpiada] 'Olympiad'
g. [hiii] [hiw] 'river'
h. [asoviii] [asoviw] 'whistle'
i. [periiidii] [periwdii] 'period'
Forms in (35a-f) illustrate cases in which a primary stressed high vowel is
immediately followed by the segment [a]. In (35a-d) the high vowel bears
penul ti~ate stress and in (35e-f) the high vowel bears antepenultimate
stress. Forms in (35g-i) illustrate cases in which a primary stressed high
front vowel is immediately followed either by the segment [ii] or by the
segment [w]. In (35g-h) the high front vowel bears penultima~e stress and in
(35i) the high front vowel bears antepenultimate stress. Let us first
consider the derivation of forms in which a primary stressed hig~lvowel is
followed by the segment [a], e.g. [pia] 'sink' and Chua] 'street'. In (36)
However, such forms have not been reported in BP. That is, in forms which
present post tonic glide-vowel sequences primary stress must fall on the
nuclear position which immediately precedes the glide, i.e. [inikwa]
'iniquitous'but ~[inikwa]. For the metrical constraints imposed on complex
consonants in BP see SILVA (1992).
29 Forms which present antepenultimate stressed high back vowels followed
by a schwa have not been reported, i.e. *[manuaka].
..,.
30 Forms in which a primary stressed high back vowel is followed by a
palatal glide, e.g. [flujdii] - [fluidu] 'fluid', will be considered in Chapter
8, where governing relations between high vowels are addressed.
31 Forms such as the ones illustrated in (35g-i) - in which a primary
stressed high front vowel is followed either by [ii] or [w] - will be addressed
later in this section.
. .'
90
CHAPTER 6: PREfONIC GLIDES AND PRETONIC HIGH VOWELS
we illustrate the lexical representation of the form [pia].32

(36) / \
s w
0 R 0 R
I I I
I I I
I
I N N
I I I
I I I
X X X
I I I
I 1 I
p---4 a
1° yO
I
t
...
It
At
In (36) we have a sequence of strictly adjacent nuclear positions which is

sUbject to government. In a governing domain the governing position must be
filled with a segment which has the adequate charm value to govern the segment
associated to the governed posl hon. Let us then consider the governing
properties of the 'segments which occupy the adjacent nuclear positions in
(36) .
The far left nuclear position is filled with the positively charmed segment
[i] and the nuclear position to the right is filled with the complex charmless
segment La]. We have seen in Chapter 1 that positively charmed segments have
the property of governing complex charmless segments. Thus, the far left
nuclear position in (36) - which is filled with the positively charmed segment
[i] - has the property of governing the following nuclear position - which is
filled with the complex charmless segment [a]. That is, a left-headed
internuclear governing domain shall be established.
-<
In Chapter 5 we proposed that the establishment of a left-headed internuclear

governing domain yields a branching nucleus formation. That is, both nuclear
posi tions are associated to a single nuclear constituent. In order for a left-
headed internuclear governing relation to hold, the same segmental
requirements imposed on branching nuclei must be satisfied. In the case of
long vowels the same segmental material occupies both nuclear positions. In
the case of heavy diphthongs the far ~eft nuclear position - which is the
governor - must be filled with either a complex charmless segment or with a
positively charmed segment. The far right nuclear position - which is the
governee - must be filled with a simplex charmless segment, i.e. [i,u,~],
Let us then consider if the segments which occupy the adjacent nuclear
positions in (36) satisfy charm requirements imposed on branching nuclei. The
far left nuclear position in (36) - which is the potential governor - is
filled with the positively charmed segment [1]. The governing nuclear position
in a heavy diphthong governing domain may be filled with either a positively
charmed segment, or with a complex charmless segment. Given that the segment
32 We assume that any form which presents a primary stressed high vowel
immediately followed by a schwa (cf. (35a-f» is derived in like manner.
91
[i] is positively charmed, it satisfies charm requirements imposed on the

governing nuclear posi tion of heavy diphthongs. The far right nuclear posi tion
in (36) - which is the potential governee - is filled with the complex
charmless segment [aJ. The governed nuclear posi tion in a heavy diphthong must
be filled with a simplex charmless segment, Le. [i,(l,~J. Given that the
charmless segment raj is complex it cannot occupy the governed position in a
.~.
heavy diphthong. Therefore, a heavy diphthong cannot be formed between the
adjacent nuclear positions in (36) because the governed position would be
filled with a complex charmless segment, i.e. [a], yielding an ill-formed
representation which is illustrated in (37).
(37) 11 0 R
I - - - - - - - - _ . _ - - - - - - - - - - - - - - - - - - - - _ .__ ._------_.-
I
N
I
I \
X X X
I I I
I I I
P i a
1.0 yO
I I
I I
...+ A+
In order to prevent an ill-formed representation - such. as the one illustrated

in (37) - we propose that a sequence of adjacent nuclear positions is allowed
to occur (cf. Chapter 5). The derivation of the form [pia] I sink I is
illustrated in (38). -
(38) s word level

~ foot level
R R nuclear projection
I I
I I
0 R 0 R
,
I I
I
I
I
I
I N N
I I I
I I I
X X X
I I I
I I I
P i a
1.0 yO
I I
1+ 1+
... A
In (38) we have a sequence of adjacent nuclear positions where the~ar left

one has the property of governing the nuclear position to the right. A left-
headed internuclear governing domain should be established. The establishment
of a left-headed internuclear governing relation yields a representation in
which both nuclear positions are associated to the same constituent, i.e. a
heavy diphthong. However, the establishment of a left-headed internuclear
governing relation between the adjacent nuclear positions in (38) would yield
92
CHAPTER 6: PRETONIC GLIDES AND PRETOHIC HIGH VOWELS
an ill-formed representation (cf. (37». Thus, in order to prevent this

phenomenon we propose that a sequence of adjacent nuclear posi tions is allowed
to occur - as illustrated in (38). Let us now consider the derivation of forms
in which a primary stressed high vowel is followed by either the segment [liJ
or by the segment [wJ (cf. (35g-h». These forms are reproduced in (39).
(39) a. [hili] [hiwJ 'river'

b. [asoviliJ [asoviwJ 'whistle I
c. [perilidli] [periwdliJ 'period'
Forms on the left in (39) illustrate cases in which a primary stressed high
front vowel is followed by a high back vowel, i.e. eft]. Forms on the right in
(39) illustrate cases in which a primary stressed high front vowel is followed
by a back glide, i.e. [w]. Let us consider the derivation of forms illustrated
in (39). In (4~) we present the lexical representation of the forms [hiliJ -
[hiw] , riverI.
(40) / \
s w
O' R 0 R
I I
1 1
N N
x x X
I I I
I I I
h i li
1° ~
I
1+
~
In (40) we have a sequence of strictly adjacent nuclear positions which is

subject to government. In order for a governing relation to hold, the
governing nuclear position must be filled with a segment which has the
adequate charm value to govern the segment associated to the governed
position. The far left nuclear position in (40) is filled with the positively
charmed segment [iJ and the nuclear position to the right is filled with the
simplex charmless segment eli]. We have seen that positively charmed segments
have the property of governing simplex charmless segments (cf. Chapter 1).
Therefore, the far left nuclear position in (40) - which is filled with the
positively charmed segment [i] - has the property of governing the nuclear
position to the right - which is filled with the simplex charmless segment y.
A left-headed internuclear governing domain shall be established.
governing domain yields a branching nucleus formation where the same segmental
requirements imposed on branching nuclei must be satisfied. The far left
33 We assume that any form which presents a primary stressed high front
vowel immediately followed by a high back vowel or a back glide is derived in
like manner.
93
nuclear position in (40) - which is the potential governor - is filled with

the positively charmed segment [i]. The segment [i] is positively charmed and
thus it satisfies charm requirements imposed on the governing nuclear position
of heavy diphthongs. The far right nuclear position in (40) - which is the
potential governee - is filled with' the simplex charmless segment y. The
governed nuclear position in a heavy diphthong must be filled with a simplex
charmless segment, i.e. [i,u,~]. Given that the simplex charmless segment y
occupies the potential governed position (cf. (40)), charm requirements
imposed on branching nuclei are satisfied so that a left-headed internuclear
governing relation holds. This process is illustrated in (41).
(41)
a. I
I ~ b.
0 R 0 R 0 R
I I I
I I I
N N N
J. I
T I \
X X x ---) x x ~
I I I I I~ \.
I I I I
h i ii h i u
1°
I
UO 1° UO
I
I 1
...+ ...+
(4ia) illustrates the establishment of a left-headed internuclear governing

relation. In Chapter 5 we proposed that the establishment of a left-headed
. internuclear governing relation yields a branching nucleus formation. In (41b)
a branching nucleus, or more specifically a heavy diphthong governing domain,
is established. In (4lb) the segment y - whose internal representation
consists of the element UO - occupies a position which is not a nuclear head.
That is, it occupies the governed position in a heavy diphthong. Given that
the element UO occupies a position which is not a nuclear head, it is
~ phonetically manifested as a glide, i.e. [w]. The form [hiw] 'river' - in
which a primary stressed high front vowel is followed by the glide [w] - is
derived.
We have seen that forms in which a primary stressed high front vowel is
followed by a glide, e.g. [hiw] , may present an alternative pronunciation in
which a sequence of vowels occurs, i.e. [hiu] (cf. (39)). In (41) we ~ave
accounted for the pronunciation of the form [hiw] , i.e. where a vowel-glide
sequence occurs. Let us now consider the derivation of the form [hiu] where
a sequence of vowels occurs. For convenience its lexical representation
presented in (40) is reproduced in (42).
94
(42) / \
s w
0 R 0 R
I I I
I I I
I
I N N
I I I
I I -I
X X X
I I I
I I I
h i 0.
1° UO
I
...1+
We have seen that a representation such as the one illustrated in (42) is
subject to government. More precisely a left-headed internuclear governing
relation holds so that a branching nucleus is formed, and a vowel-glide
sequence is manifested (cf. (41)). Notice, however, that in the process of
deriving the form [hia], i.e. where a sequence of vowels occurs, the segment
[0.] remains associated to a nuclear head position, thus appearing as a vowel.
We assume-that the derivation of the form [hia] 'river' corresponds to its
lexical representation, i.e. the structure presented in (42). Although charm
requirements are satisfied, a left-headed internuclear governing relation does
not hbld between the adjacent nuclear positions in (42) so that no vowel-glide
sequence is realized. The reasons why a left-headed internuclear governing
.relation mayor may not hold
.' between adjacent nuclear positions which are both .
filled with a high vowel (cf. (41,42)) - will be discussed in detail in
Chapter 8, where the governing relations between high vowels are addressed.
Suffices to say here that whether a governing relation holds or not, involves
sequenc
vowels.rr of adjacent nuclear positions which are both filled with high
6.6. Conclusion
In this Chapter we have considered the derivation of forms which present high
vowels and glides followed by a vowel, i.e. iV/uV sequences. The lexical
representation we propose for iV/uV sequences is illustrated in (43).
34 In Chapter 7 we will see that a left-headed internuclear governing

relation must hold between a primary stressed nuclear position filled with a
positively charmed segment other than [iJ, Le. [a,e,o,uJ, and its immediately
following nuclear position filled with a lax high vowel. Therefore, a vowel-
glide sequence must be realized. That is, unlike a form such as [hiw] - which
may also occur as [hia] - in a form such as [paw] 'stick' a vowel-glide
sequence must occur, i.e. *[paa].
95
(43) 0 R 0 R
I I
I I
N N
I I
I I
X X
I I
I I
i/u V
We have shown that a representation such as (43) is subject to the

establishment of a right-headed internuclear governing relation, since the far
right nuclear position - which is filled with any BP vowel - has the property
of governing the nuclear position to the left - which is filled with either
of the simplex charmless segments i/u. We propose that iV/uV sequences - which
are mani fested as a glide-vowel sequences - are derived from the establishment
of a right-headed internuclear governing relation, which causes the deletion
of the governed nuclear position. Thus, the segmental material which was
formerly associated to the governed nuclear position relinks to the governing
nuclear position forming a light diphthong. This process is illustrated in
(44) .
(44)
! I
I
0 R R 0 R
I I I
I I I
N N N
x x ---) x
,;>
I I ." I
I I I
i/u V i/u V
We have seen that in pretonic iV/uV sequences either a high vowel or a glide
may be phonetically manifested, e.g. [piada] - [pjada] 'joke' and [kuaAu] -
[kwaAu] 'curd' (cf. section 6.2.). We propose that pretonic iV/uV sequences
which are manifested as a sequence of vowels, e.g. [piada] and [kuaAu],
represent the case where the loss of a nuclear position is prevented. Thus,
the positively charmed ATR element is incorporated to the internal
representation of the high vowel yielding the positively charmed segments
[i,u]. This process is illustrated in (45).
(45) /I
I
! I
0 R 0 R
I I
I I
N N
x X
I I
I I
i/u V
Given that the far left nuclear position in (45) is filled with either of the
positively charmed segments [i,u] - which cannot be governed - a governing
relation does not hold between the adjacent nuclear positions. No nuclear
position is deleted and a sequence of vowels is realized.
96
We have also seen that in posttonic iV/uV sequences a glide-vowel sequence

must be phonetically manifested, e.g. [sabja] but *[sabia] 'wise' and [agwa]
but *[agfta] 'water'. Such a requirement (that a glide-vowel sequence must be
phonetically manifested in posttonic iV/uV sequences) follows from the fact
that posttonic nuclear heads must be filled with simplex charmless segments,
and that the establishment of a right-headed internuclear governing relation
in posttonic iV/uV sequences is obligatory.
Concerning iV/uV sequences where the high vowel bears primary stress, e.g.
[pia] 'sink'; [hua] 'street' and [hiu] ~ [hiw] 'river', we have seen that a
primary stressed nuclear position filled with a high vowel may govern the
following nuclear position. Thus a left-headed internuclear governing relation
may hold. We propose that the establishment of a left-~eaded internuclear
governing relation yields a branching nucleus formation.
In a branching nucleus governing domain, the far left nuclear position - which
is the head - must be filled with either a positiyely charmed segment, Le.
[a,e,i,o,u], or with a complex charmless segment, i.e. [E,~]. The far right
nuclear position - which is the governee - must be filled with a simplex
charmless segment, i.e. [i,u,~]. If the segmental requirements imposed on
branching nuclei are not satisfied when a left-headed internuclear governing
domain may potentially hold, then a sequence of adjacent nuclear positions is
. allowed to occur.
We have also considered forms in which a velar stop-back glide sequence
occurs, i.e. kw and ~. We showed that the syllabification of the back glide
in BP may correspond either to the element UO syllabified in the
representation of a light diphthong (as a consequence of the establishment of
a right-headed internuclear governing relation) or to the element UO together
with a velar stop consonant associated to a single onset position (in the
representation of a complex consonant). In (46) we illustrate the
representation of a~complex consonant.
(46) 0 R
I
I
N
x x
I
/ \ I
cUO V
We have claimed that when the back glide is phonologically interpreted as the
element lfl syllabified in the representation of a complex consonant (cf.
(46) ), it corresponds to the cases when the element UO is lexically associated
to a position other than a nuclear head. In the following Chapter we will
consider forms in which ei ther a postvocalic high vowel or a postvocalic glide ..
is realized.
35 Left-headed internuclear governing relations are discussed in detail

in the following Chapter.
97 .. ~
·: .
CHAPTER 7
POSTVOCALIC GLIDES AND POSTVOCALIC
HIGH VOWELS
7.0. Introduction
In this Chapter we consider forms which present vowel-glide sequences, i.e.
[Vj,Vw], and forms which present a sequence of vowels where the far right
member is a high vowel, i.e. [Vi,Vu]. In the first section we discuss the
phonological representation of vowel-glide sequences. In section 2 we provide
evidence that vowel-glide sequences in BP are derived from a sequence of
adjacent nuclear positions and we determine how postvocalic glides are
syllabified in BP. In section 3 we will consider the phonological behaviour
of vowel-glide sequences, and sequence of vowels in BP in relation to primary
stress. We will determine the conditions under which a vowel-glide sequence
alternates or fails to ,alternate with a sequence of vowels.
7.1. The Representation of Vowel-Glide Sequences
In this section we consider the phonological representation of vowel-glide
sequences. Glides represent the cases in which the simplex charmless segments
i/u occupy a position othe~ than a nuclear head. Therefore, when representing
vowel-glide sequences we will have a vowel occupying a nuclear head position
immediately followed by either of the simplex charmless segments [i,u]
occupying a position -other than a nuclear head. In (l) we illustrate the
potential syllabifications of vowel-glide sequences.
( 1) a. 0 ·R b. 0 R 0 R
I I I
I I I
N N N
j \
.! x x x X
I I I I
I I I I
V iju V iju
(la) illustrates a heavy~diphthong governing domain. In a heavy diphthong
governing domain the governor, i.e. the far left nuclear position, must be
filled with either a positively charmed regment, e.g. [a,e,i,o,u], or with a
complex charmless segment, e.g. [E,~]. The governed position in a heavy
1 In Chapter I we have seen that the complex charmless segments in BP are

[E,~,a]. Cases in which a complex charmless segment occupies the governing
position in a heavy diphthong, only the complex charmless segments [E,~] may
fill this position. The fact that the complex charmless segment [a] cannot
occupy the head position in a heavy diphthong, i.e. *[aj], follows from "Ia
contrainte de Kryocephalie" (cf. LOWENSTAMM (1986)). This constraint prevents
a segment that has the cold element as its head to be the head of a governing
domain (cf. Chapter 3). A heavy diphthong represents a governing domain where
the nuclear position is the head. Thus, given that the segment [a] - whose
internal representation is (A+.yo)o - has the cold element as its head, it
cannot occupy the governing position in a heavy diphthong governing domain,
98
. -~ . .,."
.: .
CRAPTER 7: POSTVOCALIC GLIDES A!D POSTVOCALIC RIGH VOWELS
diphthong must be filled with a simplex charmless segment. In (la) the

governed position of the heavy diphthong is filled with either of the simplex
charmless segments i/ii. Notice that in (la) the segments 1/ii occur in a
position which-is not a nuclear head, i.e. they occupy the governed position
in a heavy diphthong. Given that the segments i/ii occur in a position other
than a nuclear head they are manifested as glides. (la) is realized as a
vowel-glide sequence, i.e. [Vj,Vw].
(lb) illustrates the case where a nuclear position filled with a vowel is
immediately followed by an onset position filled with either of the simplex
charmless segments 1/ii. According to the theory, a non-governing onset
-p6si-tlOfi'lI\ay-b1! ~·n rea~-wtth- ~i thtr-a: -n-egatl vety-e'ltarine-cr-segment-o-r'with -a-
charmless segment (cf. Chapter 1). Given that the segments i/ii are charmless,
they satisfy charm requirements for segments occurring in onset positions.
Notice that in (lb) the segments i/ii occupy a position other 'than a nuclear
head, i.e. they occur in an onset position being thus realized as a glide.
(lb) is realized as a glide-vowel sequence. 2 In the following section we will
consider the phonological interpretation of vowel-glide sequences in BP.
7.2. The Phonological Interpretation of Vowel-GlideSequences in BP

In this section we will argue that vowel-glide sequences in BP are derived
from a sequence of adjacent nuclear positions., We will also determine how
vowel-glide ~equences, Le .. postvocalic glides, are syllabified in BP.
Consider (2)._
(2) a. [erSj] 'hero'
b. [muzew] 'museum'
c. [gajta] I tin fife I
d. [kawza] 'cause'
e. [kajpira] 'bumpkin'
f. [bawru] 'Bauru'
Le. *[aj].
2 In this section we are considering cases in which a vowel-glide
sequence occurs. Therefore, in (lb) the final nuclear position is empty so
that a vowel-glide sequence is realized. Whenever an onset position filled
with a simplex charmless segment is immediately followed by a filled nuclear
posi tion we have intervocalic glides, i.e. [VjV, VwV]. Intervocalic glides are
addressed in Chapter 8.
3 It is important to mention that in most BP dialects there is a process
of I-vocalization. This process yields a lateral consonant that is
phonetically manifested as a back glide either when it occupies a rimal
position, e.g. /altiira/ --> [awtura] 'height', or when it is followed by a
final empty nucleus, e.g. /saI0/ --) [saw] 'salt'. The forms we consider in
this work which present back glides exclude those which are related to the
process of I-vocalization.
99
Forms in (2) illustrate instances of vowel-glide sequences. 4 Concerning the

phonological interpretation of forms in (2), one could posit two hypothesis.
In one case it would be assumed that the glide is lexically present, whereas
in another case it would be assumed that the glide is derived from its
corresponding high vowel. Therefore, we have to determine whether postvocalic
glides in BP are lexically present, or whether they are derived from their
corresponding high vowels. In the following pages we will argue that the
proposal that postvocalic glides are lexically present in BP cannot be
sustained.
We have seen in section 7.1 that postvocalic glides may represent either the
cases where a simplex charmless segment is syllabified in the governed
~po.sLtion.._oL_a... hea~~_.diphthong . (c.L.Jl"al) t . . O.L .t.he.....c.a.s...e~ ...'ll!ler~. JL. simRie;
charmless segment occupies an onset position (cf. (lb)). Let us first assume
that postvocalic glides in BP are lexically present and that they correspond
to a simplex charmless segment syllabified in the governed posi tion of a heavy
diphthong. According to this proposal a form such as [muzew] 'museum' is
syllabified as in (3).
(3) 0 R 0 R
I I I I
I I I I
I
I
I N I N
I I I I
I I I I \
X X X X X
I I I I I
I I I I I
m u z e u
Heavy diphthongs present two nuclear positions associated to the same nuclear
constituent. In metrical terms a heavy diphthong counts as a single nuclear
head. Evidence that vowel-glide sequences in BP do not correspond to the
structure of a lexical heavy diphthong - as illustrated in (3) - comes from
the behaviour of vowel-glide sequences wi th respect to primary stress. Primary
stress in BP may be either final, penultimate or antepenultimate, e.g. [para]
'Para', [kaza] 'house I and [silaba] I syllable' (cf. Chapter 2). If vowel-glide
sequences were lexically present in BP and they corresponded to the structure
of a heavy diphthong - as illustrated in (3) - then we would expect to find
antepenultimate stressed vowel-glide sequences, e.g. [nawtiku]. This is
because a heavy diphthong metrically counts as a single nuclear head. Thus,
a form such as [nawtlku] would illustrate an instance of antepenultimate
stress. However, forms such as [nawtiku], i.e. which present antepenultim~te
stressed vowel-glide sequences, are not representative of the BP lexicon.
4 Forms which present pretonic postvocalic glides, e.g. [kajpira] (cf.

(2e-f)), present an alternative pronunciation in which a sequence of vowels
occurs, i.e. [kaipira]. The conditions under which a postvocalic glide or a
postvocalic high vowel is manifested are discussed in the following section.
S It is worth mentioning that although we do not find many forms
presenting antepenultimate stressed vowel-glide sequences, e.g. [nawtiku],
forms presenting antepenultimate stressed branching rimes are very recurrent
in BP, e.g. [plastlku] 'plastic', [sShdidu] 'sordid'. Forms presenting
antepenultimate stressed branching rimes, e.g. [plastlku], show that BP allows
100
...
CHAPTER 7~ POSTVOCALIC GLIDES AND POSTVOCALIC HIGH VOWELS
There are few forms which present antepenultimate stressed vowel-glide

sequences, e.g. [nawtikli] 'nautical' and [fahmasewtikli] 'pharmaceutical'.
These forms apparently support the hypothesis that vowel-glide sequences are
. lexically present, i.e. they would illustrate instances of antepenultimate
stressed vowel-glide sequences. Nevertheless, the forms which present
antepenultimate stressed vowel-glide sequences are quite rare in BP and so we
. cannot assume that vowel-glide sequences are lexically present. We propose
that forms which present antepenultimate stressed vowel-glide sequences, e.g.
[nawtikli] and [fahmasewtikli], are analysed as exceptional. They belong to a
small set of form~ which allow primary stress to fallon the fOfrth-to-last
syllable, e.g. [tEkinika] 'technical' and [hitimika] 'rhythmic'. Thus, forms
such as [nawtlkli] and [fahmasewtikli] - whose lexical representations we assume
are /nalitikli/ and /fahmaselitikli/ respectively - are exceptionally stressed on
the fourth-to-Iast syllable.
There is further evidence that vowel-glide sequences are not lexically present
in BP, and that glides do not correspond to a simplex charmless segment
syllabified in the structure of a heavy diphthong (cf. (3». This evidence
comes from the phonological behaviour of languages which have lexical heavy
diphthongs. In such languages there is usually a contrast between short and
long vowels, and there are also segm~ntal restrictions concerning the vowel
in the vowel-glide sequences which correspond to heavy diphthongs. More
specifically, a reduced set of vowels from the language may occupy the vowel
position in the vowel-glide sequences which correspond to heavy diphthongs.
Thus, i f vowel-glide sequences in BP corresponded to lexical heavy diphthongs
- as illustrated in (3) - there would be constraints concerning which vowels
could prec~de the glide. Consider (4).
(4) a. gaj ta l 'tin fife'
b. OIEi kU ] 'oleic'
c. sej a~ 'sect'
d. m6jta 'thicket'
e. bez5~kii] 'benzoif'
r
f. fluj li] 'fluid'
fi:i. aw al
hubEw a]
'cause'
. 'rubella'
•
dewzaJ
periw ii]
'goddess'
~: k6wrii]
'leriod'
, eather'
antepenultimate stressed heavy syllables. Therefore, the fact that we do not

find many forms presenting antepenultimate stressed vowel-glide sequences,
e.g. [nawtikii], is not related to any constraint which does not allow
antepenultimate stressed forms in BP to present heavy syllables.
6 For reasons which are not yet clear all forms which are exceptionally
stressed on the fourth-to-last syllable are derived forms which present the
suffix -ik.
7 A form such as [flujdli] may also be pronounced as [fluidii]. In Chapter
8 we discuss these forms in detail, where the governing relations between high
vowels are addressed. .
101
Forms in (I) show that any BP vowel may precede the glide in the glide-vowel
sequences. The lack of segmental restrictions concerning which vowels may
occur in vowel-glide sequences in BP - as illustrated in (4) - provides us
wi th evidence that vowel-glide sequences in BP do not correspond to the
structure of a lexical heavy diphthong. This is because if vowel-glide
sequences corresponded to the structure of a lexical heavy diphthong, there
would be constraints on which vowels could precede the glide, which is not the
case. We have just seen that BP does not display the typical behaviour of
languages which have lexical heavy diphthongs, since there is no segmental
restriction concerning the vowels in vowel-glide sequences (cf. (4».
Furthermore, when considering BP vfwels we observe that no contrast between
short and long vowels is attested.
We have just seen that the proposal that vowel-glide sequences are lexically
present in BP and that they correspond to the structure of a heavy diphthong
(cf. (3)) cannot be sustained. Let us now consider the proposal that vowel-
glide sequences are lexically present in BP and that the glide in the glide-
vowel sequence corresponds to a simplex charmless segment syllabified in an
onset position (cf. (lb). According to this proposal a form such-'as [muzew]
'museum' is syllabified as in (5).
(5) 0 R 0 R 0 R
I I I I I I
I I I I I I
I I I
I N I N I N
I I I I I I
I I I I I I
X X X X X X
I I I I
I I I I
m u z e U
We argue that the glide in the vowel-glide sequences in BP does not correspond
to a simplex charmless segment syllabified in an· onset position - as
illustrated in (5). Evidence that postvocalic glides in BP do not correspond
to a simplex charmless segment syllabified in an onset position, comes from
the behaviour of derived forms of finally stressed vowel-glide sequences.
Recall that the morphological structure of nouns in BP corresponds to a noun
stern followed by the gender suffix, e.g. /trabaA + u/ --> [trabaAu] 'work'
(cf. Chapter 2). The gender suffix corresponds to a nuclear position which may
be filled with either [a,u,i], e.g. [kaza] 'house', [ladu] 'side' and [b5di]
'goat'. In derived forms the noun stem is followed by one or more suffixes
which are then followed by the gender suffix nuclear position, e.g. /trabal
+ ad + a/ --> [trabaAada] _'worked' and /trabaA + ad + or + a/ -->
[trabaAad6ra] 'worker (F)'. Consider the derived forms in (6).
8 The only segmental restriction we find in vowel-glide sequences in BP

is that the segment [:>J cannot precede the back glide, Le.*[:>w]~_
9 We refer the reader to Chapter 2 where we provide .evidence thatBI)' has
no long vowels.
102
C}APTER 7: POSTVOCALIC GLIDES AND POSTVOCALIC HIGH VOWELS
(6) a. [muzew] 'museum'

[muze:,lo.gii] 'specialist in museums'
b. [iskahsEw] 'swell (of waves)'

[iskahseado.] 'swelled'
c. [er:,j] 'hero'
[eroismo.] 'heroism'
In each pair of words illustrated in (6) the first form corresponds to an
underived noun, and the second one corresponds to one of its derived forms.
-- --All--the-urrderlved--rrOUIls-in To) ,-f ~-e---:--[muzewL- [TsKansE-wT;- [erSn,--p-resent-a
final stressed vowel-glide sequence. Notice that in the derived forms, i.e.
[muze:,lo.gii], [iskahseado.], [eroismo.], no vowel-glide sequence occurs.
In derived forms the suffixes are added directly to the noun stem~, i.e. the
gender suffix is suppressed, and is added after the las~suffix. The derived
forms illustrated in (6) are evidenceq that glides in word-final vowel-glide
sequences do not belong to the noun stem. Instead, the glides correspond to
the gender suffix nuclear position. This is because in all derived forms in
(6) the word-final glide of the.underived noun is suppressed when a suffix is
added to form a derived form. Therefore, the fact that word-final glides
correspond to the gender suffix leads us to assume that vowel-glide sequences
in BP do not corre,~ond to ,the segments i/O. lexically syllabified in an onset
position (cf.(5». -
In th.e preceding pages we have provided evidence that the glide in vowel-glide
sequences in SP does not represent a simplex charmless segment lexically
associated to a position other than a nuclear head. That is, the glide in
vowel-glidesequences does not represent either a simplex charmless segment
lexically syllabified in the governed position of a heavy diphthong (cf. (3»,
nor does it represent a simplex charmless segm~ntlexical+y syllabified in an
onset position (cf. (5». Therefore, postvocalic glides are not lexically
present in SF.
We propose that postvocalic glides in SP are derived from a sequence of
strictly adjacent nuclear positions, where the far left nuclear position is
filled with any BP vowel and the far right one is filled with either of the
simplex charmless segment i/O.. In (7) we illustrate the lexical representation
of postvocalic glides, i.e. vowel-glide sequences, in. BF.
10 The reasons why the gender suffix nuclear position is phonetically

manifested as a glide rather than as a vowel will be presented in the
following section. Below we illustrate derived forms which present the
suffixes illustrated in (6), i.e. -~lug, -ad, -ism, where the gender suffix
surfaces as a vowel rather than as a glide:
[futurO.] 'future' [futur:,lo.gii] 'future teller I
[uzo.] 'use' [uzadG] 'used'
[komlina] 'commune' [komunismo.] 'communism'
103
-J.,
(7) 0 R 0 R
I I
I I
N N
I I
I I
X X
I I
I I
V 1/11
(7) illustrates a sequence of strictly adjacent nuclear positions. Recall that
adjacent nuclear positions are subject to government (cf. Chapter 1). In a
governing domain the governing position must be filled with a segment which
has the adequate charm value to govern the segment associated to the governed
_pos.ition.__Let _us.then __consi.deL.the ..g.ov.erning_prop-ert i.es Qf.the .segment.s w_htcll
occupy the adjacent nuclear positions in (7).
The far left nuclear position in (7) is filled with any BP vowel and the far
right nuclear posi tion is filled with a simplex charmless segment. Recall that
simplex charmless segments may be governed by any other nuclear segment (cf.
Chapter 1). Therefore, the far left nuclear position in (7) - which is filled
with the vowel - is the governor and the nuclear position to the right - which
is filled with either of the simplex charmless segment i/11 - is the governee.
A left-headed internuclear governing domain is established.
governing domain yields a branching nucleus formation. That is, the governed
nuclear position dissociates from its nuclear constituent and relinks to the
governing nucleus. This process is illustrated in (8).
(8) .
a. Government b. Constituent c.Branchinq nucleus
dissociation under formation
government
! I
I !
0 R 0 R 0 R 0 R 0 R
I I I I
I I I I
N N N N N
I J. I
I "T I \
~ X ~ X ~ X
I I I I I I.
I 1 I I I I
V i/11 V. i/11 V i/11
In a left-headed internuclear governing domain the segments which occupy the

adjacent nuclear positions must satisfy charm requirements imposed on
. branching nuclei. That is, the governing nuclear position must present either
a positively charmed segment or a complex charmless segment. The governed
104
.: .
position must be filled wi1~ either a simplex charmless segment or it must be

an empty nuclear position.
In (8) the governed nuclear position is filled with a simplex charmless
segment, Le. i/u, so that charm requirements imposed on the governed position
of branching nuclei are satisfied. The governing nuclear position in (8) may
be filled with any positively charmed or complex charmless nuclear segment
from BP, Le. [a,e,i,o,u,E,::>]. 'rhus, charm requirements imposed on the
governing position of branching nuclei are also satisfied.
According to our proposal presented in (8) vowel-glide sequences in BP are
----Ie.-e-ri-vea--f-Fem-the est-ab-H-s-R:me-at-e.f-~-t~-;i.n-t-e-:r.IluG-1-ea-rgoverning
relation, which yields a branching nucleus formation. In the following section
we will consider the phonological behaviour of BP vowel-glide sequences and
BP sequences of vowels in relation to primary stress. We will see that in some
environments a vowel-glide sequence may alternate with a sequence of vowels,
i.e. [kajpira] [kaipira] 'bumpkin', while in other environments a vowel-
H
glide sequence must be realized, e.g. [gajta] 'tin fife' but *[gaita]. We will
determin the conditions under which a vowel-glide sequence alternates or fails
to alternate with a sequence of vowels.
7.3. Postvocalic Glides and Postvocalic High Vowels

Consider forms 'in (9):
(9) a. [erSj] 'hero'

b. [papaj] 'daddy'
c. [paw] 'stick of wood'
d. [mlizew] 'museum'
e. [gajta] 'tin fife'

f. [m6jta] I thicket' ~
g. [kawza] 'cause'
h. [hubEwla] I rubella'
L [kajpira] [kaipira] 'bumpkin'

j. [depojmetu] [depoimetu] 'deposi tion'
k. [bawrli] [baurli] 'Baurli'
l. [hewniCiw] [heuniCiw] 'meeting'
(9a-d) illustrate forms in which a primary stressed vowel is followed by a

glide occurring word-finally. (ge-h) illustrate forms in which a penultimate
11 When the governed position in a left-headed internuclear governing

domain is empty, the branching nucleus corresponds to the representation of
a long vowel. This is where the governing segment - which occupies the far
left nuclear position - spreads over the governed position. In this section
we are considering left-headed internuclear governing domains where the
governing position is filled with a vowel and the governed position is filled
with a simplex charmless s~gment (cf. (8»; so that a vowel-glide sequence is
manifested rather than a long vowel.
105
stressed vowel is followed by a glide. (9i-l) illustrate pretonic postvocalic

glides. In the latter cases the posttonic glide may also be manifested as a
high vowel (cf. forms on the far right column in (9i-l)). In the following
pages we will consider tfie derivation of the forms [kajpira] - [kaipira]
'bumpkin'. Consider (10).
(10) / \
s w
ORO RI 0I R 0 R
I I I I I
I I N' I I I I
II ~I I
I
II ~I II ~I
X X X X X X X
I I I I I I I
I
~--l
I .l I
P t-.-r-+---------- ----
A+ 1°
(10) illustrates the lexical representation of the forms [kajpira]· and

[kaipira] 'bumpkin'. Let us first consider the ·derivation of the form which
presents a vowel-glide sequence, i.e. [kajpira]. We propose that vowel-glide
sequences in BP are derived from a left-headed internuclear governing relation
yielding a heavy diphthong formation (cf. (8)). In (10) we have a sequence.of
adjacent nuclear positions where the far left nuclear position - which is
filled wi th the posi ti vely charmed segment [a] - has the property of governing
the nuclear position to the right - which is filled with the simplex charmless
segment [i]. Therefore, a left-headed internuclear governing domain is
establ ished between the adjacent nuclear positions in (10), and a heavy
diphthong is formed. This process is illustrated in (11).
(11 )
a. / \ / \
I w
I 1 s s W
0 R 0 R 0 R 0 R b. 0 RI 0 R 0 R
I I I I I I I I t I I I
I I I I I I
N I I I I
I
N N I N N I N I N
II I
I
~
I
I
II I
I
II I
I
I
I
I
I \ II I
I
II I
I
X
I
X
I
X
I
X
I
X
I
X
I
xI --) x
I
xI xI x
I
xI xI X
I
J. I I I J.
I
k
I
a 1 p i r
I
a
I
k a 1
I
p !
1
I
r
I
a
A+ 1° A+ 1°
In (11) a left-headed internuclear governing domain is established yielding

a heavy diphthong formation. The form [kajpira] (which presents a vowel-glide
sequence) is derived, and the metrical structure is built onto the projection
of nuclear heads. This is illustrated in (12).
-4'
12 Recall that primary stress is lexically assigned (cf. Chapter 2).
Secondary and word level stress are assigned after phonological processes
apply. The final metrical structure is built onto derived forms. The
phonological expressions which appear below the segments associated to the
nuclear positions in (10) correspond to the internal representation of these
segments.
106
"' "
(12)
w
/'- s word tree level
~
I
I
S S w foot level
I I I
J. J. J.
I I I
I I I
0 R 0 R 0 R
I f I I I
I f I I I
I I
I N I N N
I I I
I I \ I
X X X X X X X
I I I I I
I I I I I
k a i p i r a
A+ 1°
Let us now consider the derivation of the form [kaipira] in which a sequence
of vowels (rather than a vowel-glide sequence) occurs. We have to explain why
in the process of deriving the form [kaipira] (from the lexical representation
illustrated in (10)), the pretonic nuclear posi tion filled wi th the high vowel
remains associated to a nuclear head position. This occurs rather than the
pretonic nuclear position filled with the high vowel being incorporated into
the structure of a branching nucleus (cf. (11)). We propose that in order to
allow a preton{c high vowel to occur as a nuclear head, the ATR element ++ is
added to the internal representation of th~ simplex charm ss segments [i,u].
Thus, yielding the positively charmed segments [i,u]. This process is 11
(13) / \ / \
s w s w
a.O R 0 R 0 R 0 R b. 0 R 0 R 0 R 0 R
. I
I
I
I
I
I
I
I
I
I ,
I I
I
I
I
I
I
I
I
I
I
I I I
I N N N N I N N N I N
I
I
I
I
I
I ,
I I
I
I
I
I
I
I
I
I
I
I
I
I
I
X X X X x· x X --> x X X X X X X
I I I f I I I I I I I
I I I I I I I I I I I
k a i p i r a k a i p i r a
A+ 1° A+ 1° I
1+
+
In (13b) we have a sequence of adjacent nuclear positions where both nuclear
positions are filled with positively charmed segments, i.e. [i,a]. Given that
positively charmed segments cannot be governed (cf. Chapter 1), no governing
relation holds between the adjacent nuclear positions in (13). Thus, a
sequence of nuclear position is allowed to occur and a sequence of vowels is
13 Recall that pretonic nuclear heads must be filled with positively

charmed segments (cf. Chapter 3). The segments [i,u] are positively charmed,
50 that charm requirements imposed on pretonic nuclear heads are satisfied.
107
realized. In (14) we illustrate the final derivation of the form [kaipira],

after the metrical structure is built.
w~s
(14 )
word tree level
1---- :~
s w s w foot level
I I I I
..!. I 1 I
I I I I
1 I I I
0 R 0 R 0 R 0 R
I I I I I I
I 1 I I I 1
I I
I N N I N N
I I I I I I
I I I I I 1
X X X X X X X
I I I I I I
I I I I I I
k a i p i r a
A+ 1°
I
1+
~
In the preceding pages we have accounted for forms in which a pretonic

postvocalic glide alternates with a pretonic postvocalic high vowel, i.e.
[kajpira] - [kaipira]. Posttonic 'postvocalic glides cannot al temate with
their corresponding high vowel, e.g. [muzew] 'museum' but *[muzeu] (cf. (9a-
h)). The fact that this does not occur follows from charm constraints imposed
on BP nuclear segments and the establishment of a left-headed internuclear
governing relation. Posttonic nuclear heads in BP must be filled wi h
charmless segments (cf. Chapter 3). Therefore, a posttonic nuclear position
filled with either of the simplex charmless segments ifu - will be governed
by an immediately preceding nuclear position filled with any BP vowel. A left-
headed internuclear governing relation always holds between a posttonic
nuclear position filled with a simplex charmless segment, i.e. ifu, and the
immediately preceding nuclear position which may be filled with any BP vowel.
A posttonic postvocalic glide occurs.
7.4. Conclusion
In this Chapter we have considered the derivation of forms which present
postvocalic glides, and forms which present postvocalic high vowels. We have
seen that whether a postvocalic glide or a postvocalic high vowel is
manifested depends on whether or not a left-headed internuclear governing
relation is established.
Cases in which a postvocalic glide occurs illustrate instances where a left-
headed internuclear governing relation is es..t,ablished, yielding a heavy
diphthong formation. Cases in which a postvocalic high vowel occurs illustrate
instances where a high vowel acquires the ATR element, appearing thus as a
positively charmed segment, Le. [i,u]. A sequence of adjacent nuclear
positions is allowed to occur since a positively charmed segment cannot be
governed. In the following Chapter we consider forms which present
intervocalic glides and forms which present a sequence of high vowels.
108
CHAPTER 8
INTERVOCALIC GLIDES AND SEQUENCES
OF HIGH VOWELS
8.0. Introduction
In this Chapter we will analyse forms in which a glide is flanked between two
vowels, e.g. [meja] 'half', and forms presenting sequences of high vowels. In
the first section we determine how intervocalic glides are phonologically
interpreted in BP. In the remaining sections we consider forms which present
sequences of high vowels, determining which of the high vowels in the sequence
may be phonetically manifested as a glide.
8.l. Intervocalic Glides

Consider forms in (1).
(1) a. [saja] 'skirt'
b. [idEja] 'ideal
c. {klija] 'calabash gourd'
d. . [mejil] 'middle'
e. [ap6jil] 'support'
f. [gojaba] 'guava'
g. [maj6] 'bathing suit'
h. [bajlika] 'small tavern'
i. [fej6zil] 'uglyish l
j. [bajoneta] 'bayonet'
Forms in (1) illustra\e intervocalic glides. In (la-e) the intervocalic glide
occurs posttonically. In (If-j) the intervocalic glide occurs in a pretonic
position. In the following pages we will consider how intervocalic glides are
phonologically interpreted in BP. The simplest hypothesis one can posit is to
assume that intervocalic palatal glides represent cases in which the segment
1 - whose internal representation consists of the element 1° - occupies an
onset position.·
We have seen that when the element 1° occupies a posi tion other than a nuclear
head it is phonetically manifested as a palatal glide. Thus, in cases where
the element 1° occupies an onset position - which is not a nuclear head
position - it appears as a palatal glide. According to this proposal, a ..,.
form
1 In our research we found that forms which present intervocalic back

glides are all names whose origin comes from Amerindian languages. Considering
they are not representative, we did not incorporate them in our data. The data
presenting intervocalic back glides we found are: [mawa] 'Maua (town) I,
[ananidewa] 'Ananindeua (town)' [piawi] " [pjawi] 'Piaui (state) and [kawe]
I
'Caue (cement factory)'.

109
such as [saja] 'skirt' will have the following representation.

(2) 0 R 0 R
I I I
I I I
N I
I N
I I
I I
X X X X
I 1 I I
I I I I
S a 1° a
If intervocalic palatal glides represent the element 1° syllabified in an
onset position - as illustrated in (2) - we can expect to find antepenultimate
stressed forms which present posttonic intervocalic glides. That is, [lakajii]
would hypothetically be a possible form. Thus, in a form such as [lakajii]
stress would fallon the third-to-Iast nuclear position, corresponding to the
antepenultimate stress pattern~ Forms such as [ezitii] 'success' and [lakaju]
would represent the same stress pattern. In the form [ezitu] the final onset
is filled with the segment 1 and in the form [lakajii] the final onset is
filled with the segment i.
However, forms such as *[lakaju] do not occur in BP. Whenever an intervocalic
glide occurs preceding the gender suffix vowel, i.e. the final unstressed
vowel of a noun, primary stress falls on the vowel which precedes the glide,
i.e. [lakaju] 'lackey'. The lack of forms such as *[lakaju] seems to give us
evidence that intervocalic glides do not represent the element 1° syllabified
in an onset position (cf. (2)). This is because if . palatal glides were
syllabified in an onset position nothing would pr~vent stress from falling on
the second-to-Iast nuclear position which precedes the glide, i.e. *[lakajii] .
should be a possible form.
Nevertheless, for reasons which are not yet clear, BP does not allow palatal
consonants to occupy the final onset position in antepenultimate stressed
nouns. That is, forms such as *[maniAa], *[galifia], *[kurii~a] and *[bulaSa]
are not possible in BP. Whenever a palatal consonant occupies the final onset
posi tion in a noun, primary stress falls on the nuclear position which
precedes the palatal consonant, e~g. [manHa] 'shackle', [galifia] 'hen',
[kuru~a] 'owl', [bulMa] ,biscuit' .
We may then propose a constraint according to which palatal consonants cannot

occupy the final onset position in antepenultimate stressed nouns in BF.
According to this constraint forms such as *[manHa], *[galifia], * [kurii~a] and
*[bulaSa] are prevented from occurring. Notice that this constraint may also
account for the lack of forms which present posttonic intervocalic glides,
e.g. *[lakajii], if we assume that the palatal glide corresponds to the element
1° syllabified in an onset position (as proposed in (2)). This is because the
2 The strong R - which we have transcribed as [h] - is also not allowed

to occur in such a position, i.e. *[sigahu] is not a possible form in BP.
Whenever the strong R occupies the final onset position in a noun, primary
stress falls on the nuclear position which immediately precedes it, Le.
[sigahii] 'cigarette'(cf. Chapter 2).
110
palatal glide would correspond to a palatal consonant and therefore it could

not occupy the final onset position in antepenultimate stressed nouns.
Therefore, the proposal that palatal glides correspond to the element 1°
syllabified in an onset position - as illustrated in (2) - appears to be
. correct. The lack of forms which present posttonic intervocalic glides, e.g .
• [l~kajii], would reflect a general constraint according to which palatal
consonants cannot occupy the final onset position in antepenultimate stressed
nouns in BP, i.e. *[manila], *[g~lifia], *[kurii~a], *[bulaSa], .[l~kajii] are
not possible forms.
On the other hand palatal consonants can occupy the onset position immediately
following antepenultimate stresfed vowels, e.g. [kanarnii] 'hemp', [mESikii]
'Mexico' and ~fri~idii] 'frigid'. Therefore, if intervocalic glides represent
the element I syllabified in an onset position ~ as illustrated in (2) - we
expect to find antepenultimate stressed forms where the glide immediately
follows the primary stressed vowel, i.e. [g6jaba] would be a poSsIble form.
This is because other palatal consonants can occupy ·the onset position
immediately following antepenultimate stressed vowels, e.g. [fri~idii]
, frigid'. Then, nothing would prevent the palatal glide - which would
correspond to the element 1° syllabified in an onset position - from occupying
such a position.
Nevertheless, forms where intervocalic palatal glides occur immediately after
an antepenultimate stressed vowel, Le. *[g6jaba], do not occur in BP.
Whenever an intervocalic glide occurs preceding the final nuclear position of
a noun stem, primary stress falls on the ruclear position which immediately
follows the glide, e.g. [gojaba] 'guava'.
Thus, the lack of forms where intervocalic glides occur immediately after an
antepenultimate stressed vowel, Le. *[g6jaba], provides us with evidence that
the phonological representation of intervocalic palatal glides does not
correspond to the element 1° syllabified in an onset position. This is because
if palatal consonants may occupy the onset position which immediately follows
antepenultimate stressed vowels, Le. [fri~idii], nothing would prevent us from
finding antepenultimate stressed forms where the intervocalic glide followed
antepenultimate stressed vowels. That is, *[g6jaba] should be a possible form
- which is not the case.
If intervocalic palatal glides do not represent the element 1° syllabified in
an onset position - as illustrated in (2) ~ we have to determine what the
phonological interpretation of palatal glides is. Recall that palatal glides
3 In our research we have not found forms in whi~ a palatal lateral

occupies the onset position immediately following an antepenultimate stressed
vowel, e.g.[palasii].
4 Recall that the morphological structure of nouns in BP consists of a
stern followed by the gender suffix. Therefore, in a form such as' [gojaba]
'guava the final nuclear posi tion of the noun stern is the one which precedes
I,
the gender suffix vowel, i.e. the final vowel of the noun (cf. Chapter 3).
111
represent cases in which the element 1° occupies a position other than a

nuclear head. In (3) we illustrate some other possible representations of
intervocalic glides.
(3) a. 0 RO
1
N
,
N
R b. 0 R0
N
I
I
R
I
I
N
I
l\ I
I
I
X X X X X
I I I I
\
_ _.....
I
V 1°
I I
V V
I
It V
,- ... " .. , ..•..._.-......._.. .. _. -., .. _--- .. ', .....
In (3a) we have two nuclear positions associated to the same nuclear

constituent, i.e. a heavy diphthong governing domain. The element 1° occupies
the governed position in the heavy diphthong and is phonetically manifested
as a palatal glide. The heavy diphthong governing domain in (3a) is
immediately followed by a nuclear position which is filled with a vowel.
Therefore, (3a) is phonetically manifested as a vowel-glide-vowel sequence.
In (3b) we have a nuclear position filled with a vowel which is immediately
followed by a light diphthong. The element 1° occurs as the left member of the
segmental sequence in a light diphthong (which is not a nuclear head
position). Thus, it is realized as a palatal glide. In (3b) a vowel-glide-
vowel sequence is phonetically manifested.
80th representations in (3) correspond to vowel-glide-vowel sequences and are
therefore potential representations for intervocalic glides. We have proposed
in the preceding Chapters that both heavy and light diphthongs in BP are
derived from a sequence of adjacent nuclear positions (cf. Chapters 6-7). That
is, glides are derived from a nuclear posi tion fi lIed wi th either of the
simplex charmless segment i/u whose internal representation consists of the
elements 1° and UO respectively.
We can assume that either of the representations illustrated in (3) correspond
to the structure of vowel-glide-vowel sequences in BP. If this is the case,
then the lexical representation of vowel-glide-vowel sequences presents a
sequence of three consecutive nuclei. This is because according to our
proposal heavy and light diphthongs are derived from a sequence of adjacent
nuclear positions.
The assumption that intervocalic palatal glides are derived from a nuclear
position filled with the element 1° accounts for the distribution of primary
stress in forms which present intervocalic glides. We have seen that BP does
not allow for antepenultimate stressed forms in which a posttonic palatal
glide occurs, Le. *[lakajii] and *[g6jaba]. The lack of antepenultimate
stressed forms presenting'posttonic intervocalic glides is accounted for by
assuming that the intervocalic glide corresponds to the element 1° lexically
associated to a nuclear head position.
Thus, in a form such as /lakaiu/ --) [lakaju] 'lackey' primary stress falls
on the nuclear position which immediately precedes the glide, because
otherwise stress would fall on the fourth-to-last nuclear position, Le.
112
*[lakajii], which is not possible in BP. In a form such as /goiaba/ -->

[gojaba] 'guava' primary stress falls on the nuclear position which
immediately follows the glide, because otherwise stress would fallon the
fourth-to-Iast nuclear position, i.e. *[g6jaba], which is not possible in BP.
If the proposal that intervocalic palatal glides are derived from a nuclear
position is correct, we can expect to find forms where an intervocalic high
vowel bears primary stress, e.g. [baia] 'bay' Indeed, this is exactly the
case. 5
"_. --.--.-- -- ---~- ··-In~neprecrnngpages-we-1lave--VLop1lS-eli-·-t-h-at ·-tnt'e-rvo·c·a-l-rc-pa-l-ata-l-gi-i-d-e-g- -"---~
correspond to either the structure of a heavy diphthong followed by a nuclear

position filled with a vowel (cf. (3a», or to the structure of a light
diphthong preceded by a nuclear position filled ~ith a vowel (cf. (3b).
According to our proposal, light and heavy diphthongs are bot~ derived from
a sequence of strictly adjacent nuclear positions (cf. Chapter 6 and 7).
Therefore, if intervocalic glides correspond to either structure illustrated
in (3) we have to assume that in its lexical representation the palatal glide
corresponds to the element 1° associated to a nuclear position.
In the following pages we will derive the form [saja] 'skirt' assuming that
any form which presents intervocali,c glides is derived in like manner.
Consider the lexical representation of the form [saja] 'skirt' in (4).
(4) 0 R 0 R 0 R
I I I
I I I
N N N
I I I
I I I
X X X X
I I I I
I I I I
S a i a
At 1° yO
\i'-
~
5 We did not find forms in which an intervocalic high vowel bears

antepenultimate stress, Le. *[baiaka) or *[batiaka) . Nevertheless, fot reasons
which are still obscure, BP does not allow for antepenultimate stressed high
vowels to be preceded by a vowel, i.e. *[baikaka) or *[batikaka). Apparent
counter examples to this constraint involve forms that present analytic
morphology, more specifically, forms which present the analytic suffix -isim.
In these forms an antepenultimate stressed front high vowel may be immediately
preceded by another vowel, e.g. [feisimii] 'uglies1:'. As we have shown (cf.
Chapter 3), in forms which involve analytic morphology, the noun and the
suffix are analysed as independent morphological domains. Therefore, given
that forms which present the analytic suffix -isim involve independent
morphological domains, they are only apparent counter examples to this
constraint. Forms which present antepenultimate stressed high back vowels
immediately preceded by a vowel have not been found, Le. [natitikii] (cf.
[nawtikii] 'nautical').
113
In the lexical representation illustrated in (4) we have a sequence of three

consecuti ve adjacent nuclear posi tions which are subject to goverrunent. Recall
that goverrunent is a binary relationship between two strictly adjacent
skeletal positions (cf. Chapter I). Given that goverrunent is a binary
relation, i.e. it involves two skeletal positions, we have to show which two
strictly adjacent nuclear positions in (4) interact in a governing relation.
In order for a governing relation to hold, one of the positions must be filled
wi th a segment which has the adequate charm value to govern the segment
occurring in the other position (cf. Chapter 1). Recall that the simplex
charmless segment i may be governed by any other nuclear segment of BP (cf.
-ehapter---l-)-; --T-herefore,----an---i-rItervocaH-c--nuci-ear--pusiti-on- -ti-l-I-ed---w±th-the--
simplex charmless segment i may be governed either by the immediately
preceding nuclear position - when a left-headed internuclear governing domain
is established - or by the immediately following nuclear position - when a
right-headed internuclear governing domain is established. In (5) we
illustrate the derivation of the form [saj.a] where the glide is understood
as belonging to_ the initial syllable.
(5)
~
I
I
0 R 0 .R 0 R 0 R 0 R
I I I I I
I I I 1 1
N N N N N
J.
T \
X X X X ---) X X X X
I I I I I .1 I
I I I I I - I ., I-
S a 1 a s a 1 a
A+ 1° vo A+ 1° vo
I 1
I I
A+ A+
(5) illustrates the establishment of a left-headed internuclear governing

domain yielding a heavy diphthong. The heavy diphthong in (5) is immediately
followed by a nuclear position filled with the segment [a], so that an
intervocalic palatal glide occurs. In (6) we illustrate the derivation of the
form [sa.ja] where the glide is understood as belonging to the final syllable.
(6) illustrates the establishment of a right-headed internuclear governing

relation yielding a light diphthong. The light diphthong in (6) is immediately
preceded by a nuclear position filled with the segment [aJ, so that an
intervocalic palatal glide occurs.
In this section we have argued that intervocalic palatal glides are derived
from a nuclear position filled with the simplex charmless segment i. We showed
that an intervocalic nuclear position filled with the simplex charmless
segment i is subject to the following processes:
ar It" is govEftIred "by -the lrt1lned1a:tely 1>tec-edl-ng "nuc1-e-ar pos-i tton-.
A left-headed internuclear governing domain is established. A heavy
diphthong is formed so that a vowel-glide sequence, e.g. [saj.a], is
realized (cf. (5».
b) It is governed by the immediately following nuclear position.
A right-headed internuclear governing domain is defined. A light
diphthong is formed so that a glide-vowel sequence, e.g. [sa.ja], is
realized (cf. (6».
In the following sections we will consider governing relations between
adjacent nuclear positions both ,filled with a high vowel.
8.2. Sequences of High Vowels

8.2.0. Introduction
In the previous Chapters we have discussed governing relations between
adjacent nuclear positions. We have shown that glide-vowel sequences in BP
correspond to the structure of a light diphthong. Light diphthongs in BP are
deri ved from the establishment of a right-headed· internuclear governing
relation, as illustrated in (7).
(7)
~
I
1
0 J{ 0 R 0 R 0 R
I
I I
I
I
I
I
J( N N N
I I
1
t x ---) x ---) x
I I
I ." II / \
vI V1 VI V1 VI V1 .,.
We have also shown that vowel-glide sequences in BP correspond to the
structure of a heavy diphthong. Heavy diphthongs in BP are derived from the
establishment of a left-headed internuclear governing relation, as illustrated
in (8).
115
Ina governing domain one of the positions is the governor (or the head) and
the other -posi tion .is thegovernee (or the complement)._ The ~proper"ty _of .a
skeletal position to govern or to be governed is given by the charm value of
the segments which are linked to the skeletal positions.
We argued that the underlying representation of high vowels in BP corresponds
to the lax high vowels 1 and y - which are simplex charmless segments (cf.
Chapter 3). We have seen that simplex charmless segments may be governed by
positively charmed segments and complex charmless segments (cf. Chapter 1).
A charmless segment governs another charmless segment if it has a complexity
greater than its governee. That is, the charmless segment which has a greater
number of elements in its internal representation is' the governor. It follows
from this that neither of the simplex charmless segments 1 and g (which have
the same degree of complexity since their internal representation consists of
the elements 1° and UO respectively) should govern the other.
However,. as we will see in the following pages, a simplex charmless segment
may govern another simplex charmless segment. Evidence that this occurs
follows from light and heavy diphthong formation in BP. That is, whether a
prevocalic glide, i.e. a light diphthong, or a postvocalic glide, i.e. a heavy
diphthong, is manifested depends on the governing properties of the simplex
charmless segments 1 and y.
8.2.1. Pretonic Sequences of High Vowels

In this section we will analyse forms presenting pretonic sequences of high
vowels. We will see that these forms may present alternative pronunciations:
either a pretonic glide-vowel sequence or a pretonic vowel-glide sequence is
phonetically manifested. Consider (9).
(9) a. [~uizadu] I judgeship'
b. [ituitivu] I intui ti ve'
c. [siumera] 'jealousy'
d. [miudeza] Ilittleness'
Forms in (9a-b) illustrate cases in which a pretonic high back vowel is
immediately followed by a pretonic high front vowel, i.e. [ui]. Forms in (9c-
d) illustrate cases in which a pretonic high front vowel is immediately
followed by a pretonic high back vowel, i.e. [iu]. Forms in (9) have also the
following pronunciations:
116
CHAPTER 8: INTERVOCALIC GLIDES AND SBQUENCES OF BIGH VOWELS
(10) a. [~wizada] [~ujzada]

b. [itwi tivO.] [itujtiva]
c. [sjumera] [siwmera]
d. [mjudeza] [miwdeza]
Forms in the left column in (10) are pronounced with a pretonic glide-vowel
sequence. In the right column a pretonic vowel-glide sequence is realized.
According to our proposal, glide-vowel sequences and vowel-glide sequences are
derived from a sequence of strictly adjacent nuclear positions (cf. (7,8».
It follows from this that the lexical representation of forms in (10) present
a sequence of strictly adjacent nuclear positions. In (11) we illustrate the
lexical representation of the forms [~wizada] - [~ujzada] judgeship' and I
[mjudeza] - [miwdeza] 'littleness' respectively.

(11 )a. / \
s w
0 R 0 R 0 R 0 R
I 1 I I I I I
1 I I I I I I
I I I
I N N I N I N
I I 1 I I
I I I I I
X X X X X X X
I I I I 1 1
1 I I I 1 I
~ a i z a d a
uo 1°
b. / \
s w
0 R 0 R 0 R 0 R
I I I 1 I I I
1 1 I 1 I 1 I
,I I I
I N N I N 1 N
I I 1 I I 1 I
I I I I I I I
X X X X X X X
I I I I 1 1 I
I I I I I 1 1
m i a d e z a
1° UO
Both adjacent pretonic nuclear positions in (11) are filled with a simplex
charmless segment, i.e. i/a. We have seen that pretonic nuclear heads must be
filled with a positively charmed segment (cf. Chapter 3). Therefore, the
pretonic nuclear positions in (11) should be filled with a positively charmed
segm~t, i.e. [i,u]. We propose that the ATR element is added to the internal
representation of one of the pretonic nuclear positions in (11) so that it is
filled with a positively charmed segment. When the far left pretonic nuclear
position in the sequence of adjacent nuclear positions is filled with a
positively charmed segment, a left-headed internuclear governing relation
holds. This is because the governing position filled with the positively
charmed segment governs the immediately following nuclear position which is
filled with a simplex charmless segment. A branching nucleus is formed. The
117
forms presenting pretonic vowel-glide sequences, i .e. [~ujzadii] and [miwdeza],

are derived in (12).
(12)
a. / \ / \
I
I ! s w s w
o R o R
I
o
I
R 0 R 0 R o R o R
I I I I I I I I I
I I I I I I I I I I I
: N N , I
N I
I N N I
I N N
I
I
I
I
I
I
I
I
I
I
I
I : \ I
I
I
I
I
I
X X X X X X X --> X x x X X X X
I I I I I I I I I I I I I
I I I I I i
~ u i z a d ii ~ u i z a d ii
b. / \ / \
I
I ! s w s w
ORO R o R 0 R o R o R 0 R
I
I
I
I
f
I I
r I
I
I
I
I
I ,
I I
J
I
I
I
I
I N N : N N I
I N I
I N I
I N
I I I I
I
I I I I \ I
I I
I
I I
X X X X X X X --> X x X X X X X
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
f
I
I
I
I
I ,
I
m i ii d e z a m i ii d e Z a
When the far right pretonic nuclear position in the sequence of adjacent
nuclear positions is filled with a positively charmed segment, a right-headed
internuclear governing relation is established. This is because the nuclear
position filled with the positively charmed segment governs the immediately
preceding nuclear position which is filled with a simplex charmless segment.
A light diphthong is formed. Forms presenting pretonic glide-vowel sequences,
i.e. [~wizadii] and [mjudeza], are derived in (13).
(13)
a. I
/ \ / \
1 I s w s w
oI , o
R R
I
o
I
R
I
0
I
R
I
0
I
R
I
ORO
I I I
R
I
I N f
I I I I I N I I I I
Ir I
I
N
I
I
II ~I II ~I II I
I
II ~I II N
I
I
X X X X X X X --> X X X X X X
I
I
~
I
1
U
I
I
i
I
I
Z
I
I
a
I
I
d
I
1
U
I
I
~
L \.1
u
I
~
I
...~ d I I
1
U
118
.: .
b. / \ / \
I
1 I S W S W
0 R 0 R 0 R 0 R 0 R 0 R 0 R
I I I I I I I
I N N I N I N I N I N I N
II I
I
I
I
II I
I
II I
I
II I
I
II I
I
I
I
I
I
X X X X X X X --> X X X X X X
I
I
m
J.
1
I
u
I
I
I
d
I
I
e z
I
I
a
I
I
m
I
I t1 \-u
I
I
d
I
I
e
I
I
z
I
I
a
1° u<>
I
1°
I
yo
- -_--i+__ ...., .,., ..... ,_. _.._..__..• _. ....1+_-_.-
--- ----_... __ ............. ,......... ,..•
When both adjacent pretonic nuclear positions are filled with positively
charmed segments, no governing relation holds (because positively charmed
segments cannot be governed). Forms presenting a pretonic sequence of vowels,
i.e. [~uizadu] and [miudeza], are derived in (14).
(14)
a. / \ b. / \
s w s w
0 R 0 R 0 R 0 R 0 R 0 R 0 R 0 R
f I I I. I I I I I I I I I
I I I I I
I N N I N I N f N N I N N
I I I I I I I I I I
I I I I I I I I I I
X X X X X X X X X X X X X X
I I I I I I I f 1 I I I
I 1 I I I I I I I I I I
~ u i z a d u m i u d e z a
11:0 1° 1° yo
I I I I
I
1+ I I
...+ ... ...+ ...+
In the preceding pages we have corrsidered the derivation of forms presenting

pretonic light and heavy diphthongs (which are derived from the establishment
of a governing relation between adjacent nuclear positions filled with high
vowels) and forms presenting pretonic sequences of high vowels. The following
pages will be devoted to the discussion of other environments where governing
relations between nuclear positions filled with high vowels are observed.
8.2.2. Posttonic Sequences of High Vowels

In this section we will consider forms in which a sequence of high vowels
occurs in the final nuclear positions in antepenultimate stressed nouns. That
is, the weak positions in the antepenUltimate stress pattern ssw. Consider
forms in (15).
(15) a. [sabju] 'wise'
b. [edifisju] 'building'
c. [tenwi] 'tenuous'
d. [k5grwi] 'congruous'
119
Forms in (15) illustrate cases in which glide-vowel sequences occur in a

position which follows primary stress. In (15a-b) a palatal glide is followed
by the high back vowel [a]. In (15c-d) the back glide is followed by the high
front vowel [i]. 6 .
We have seen that glide-vowel sequences in BP correspond to the structure of
a light diphthong, which is deriv,d from the establishment of a right-headed
internuclear governing relation. Therefore, forms illustrated in (15) -
which present glide-vowel sequences - are derived from the establishment of
a right-headed internuclear governing relation yielding a light diphthong. The
derivation of the forms [sabja] and [tenwi] - which present glide-vowel
s-e-quence-s-.;;·· i-so given _
in-(-l6-Y~-" _ _... . __.._ .
(16 ) a.
I
! I
0 R 0 R 0 R 0 R 0 R
I I I I I I r
I I 1 I I 1 I
I I
N I N N I N N
I I I
I I I
X X X X X -->x x x X
I
I ,
I I
I
I
I
I
I ,
I I
1
I
/ \
I
S a b i a s a b i a
1° nO 1° nO
b.
I
! I
0 R 0 R 0 R 0 R 0 R
I
I ,
I I
I
I
I
r
I
I
I
I
I
I
I
I I I
I N I N N N I N
I I I I I I
I I I I I I
X X X X X -->x x x X
r I I I I f f f
I I I I I I I I / \
t e n a i t e n a i
nO 1° nO rO
In the governing domain illustrated in (16a) the high back vowel y governs the
high front vowel i. In the governing domain ill~strated in (16b) the high
front vowel i governs the high back vowel y. Given this, we predict that a
left-headed internuclear governing relation may also hold between posttonic
6 Forms in (15) illustrate posttonic sequences of high vowels where one

is a high front vowel and the other one is a high back vowel. Forms which
present sequences of identical high vow~s also occur in BP. These forms will
be addressed later in this section.
7 In Chapter 6 we showed that posttonic prevocalic glides may correspond
to the element nO syllabified in the representation of a complex consonant.
In this case the glide is always preceded by a velar stop consonant. The forms
discussed in this section exclude those presenting velar stop-glide sequences.
120
CHAPTER 8: INTERVOCALIC GLIDES AND SEQUENCES OF HIGH VO~ELS
nuclear positions filled with high vowels. We have seen that the establishment
of a left-headed internuclear governing relation yields a heavy diphthong, so
that a vowel-glide sequence occurs (cf. (8». Therefore, the establishment of
a left-headed internuclear governing relation between posttonic nuclear
positions filled with a high vowel would yield vowel-glide sequences. Forms
such as [sabiw] and [tenuj] - which present posttonic vowel-glide sequences -
would illustrate cases in which a left-headed internuclear governing relation
holds between a posttonic sequence of high vowels.
Nevertheless, forms such as *[sabiw] and *[tenuj] do not occur in BP. The lack
-.
of-fo-rIlfs--pr-el;jelftili9'--p-osnbnn::--vbwel~grfde--seqllEmCeS is accounted-fC>r-bytl'fe---
fact that heavy diphthong nuclear heads must be filled either with a
positively charmed segment or with a complex charmless segment (c.f. Chapter
7) •
Posttonic nuclear heads filled with a high vowel in BP must be filled with the
simplex charmless segments 1 and y. This is to fulfil charm requirements which
define that posttonic nuclear heads must be filled with a charmless segment
(cf. Chapter 3). Being that the segments 1 and yare simplex and charmless
they are unable to head a heavy diphthong governing domain. Thus, we predict
that posttonic vowel-glide sequences should not occur in BP, i.e. [sabiw] and *
*[tenuj]. This is exactly the case.
In (16) we discussed the derivation of forms which present posttonlc sequences
of different high vowels. That is, one of the high vowels is back and the
other is front. Sequences of identical high vowels may also occur in BP.In
the following pages we will show that a posttonic sequence of identical high
vowels is realized as a single vowel. Let us then analyse sequences 0i
adjacent nuclear positions filled with identical high vowels. Consider (17).
(17) 0 R 0 R
,, ,.
I
N N
I I
I I
X X
I I
I I
i i
(17) presents a sequence of adjacent nuclear positions which are both filled
wi th the simplex charmless segment 1. Sequences of adjacent nuclear posi tions
are subject to government according to the governing properties of the
segments which occupy the adjacent nuclear positions. We have seen that a
nuclear position filled with the simplex charmless segment 1 may govern a
nuclear position filled with another simplex charmless segment, i.e. y (c.f.
(l6b». Let us assume that a nuclear position filled with t'he simplex
charmless segment 1 may also govern another nuclear position filled with the
simplex charmless segment 1. That is, the simplex charmless segment 1 governs
8 A posttonic sequence of nuclear positions where both nuclei are filled

with the simplex charmless segment y is subject to the same process discussed
for the representation in (17).
121
itself. In (18) we illustrate the establishment ~f a right-headed internuclear

governing relation yielding a light diphthong. .
(18)
! I
I
a. 0 Jt 0 R b. 0 R
I
1 I
I
I
Jf N N
I
1 I
I
I
t X ---> x
I
1 I / \
···i - -i -
ii -
In the light diphthong illustrated in (18b) we have a single nuclear position

associated to two identi.cal segments - namely a sequence of the simplex
charmless segment 1. A light diphthong where the two members of the segmental
sequence consist of the same segment is interpreted as a single vowel (cf.
KAYE (1985, 198ge)). Thus, the light diphthong illustrated in (18b) is
understood as a single short vowel, i.e. [i]. Consider forms in (19).
(19 ) a. [mulEki] 'mischievous'
b. [g6stti] 'taste'
c. [kari] 'dental cavity'
d. [vakti] 'vacuous'
(19) illustrates forms in which a posttonicsimplex charmless segment occurs.
In the following pages we will show that the posttonic simplex charmless
segments in (19a-b) are lexically associated to a single nuclear position. On
the other hand, the posttonic simplex charmless segments in (l9c-d) are
derived from a sequence of posttoniTOnuclear positions which are both filled
with either of the segments 1 or g.
In derived forms the final unstressed nuclear position corresponding to the
gender suffix is suppressed when a derivative suffix is added to the noun
stem, e.g. (trabal + til --> [trabalu] 'work' and (trabal + ad + til -->
9 In this section we are considering governing relations between

posttonic sequences of nuclear positions. We have shown that posttonic
sequences of nuclear positions are subject to a right-headed internuclear
governing relation yielding a light diphthong (c.f. (16)). On the other hand,
a left-headed internuclear governing relation yielding a heavy diphthong
cannot hold between posttonic sequences of nuclei. Therefore, in the following
pages we will restrict o~ discussion to the establishment of right-headed
internuclear governing relations in posttonic sequences of identical high
vowels.
10 Forms (19c-d) may be pronounced as [karii] and [vakuu], i.e. with a
posttonic sequence of identical vowels, in careful speech. Forms in (19a-b)
must be pronounced as [mulEki] and [g6stu] in any style.
122
[trabaAadu] 'worked'(cf. Chapter 3).11 If we consider derived forms of (19a-

b), Le. (mulEk +1) --> [mulEki] 'mischievous' and (gost + u) --> [g6sto.]
'taste', we observe that the final nuclear position corresponding to the
gender suffix is suppressed when a derivative suffix is added to the noun
stern, e.g. (mulek + ad + a) --> [mulekada] 'mischief' and (gost + oz + 0.) -->
[gost6zo.] 'tasteful'.
On the other hand, derived forms of (19c-d), i.e. [kari] 'dental cavity' and
[vako.] 'vacuous', require that the final vowel (which appears to correspond
to the gender suffix) re~,ins in derived forms, e.g. [kariado.] 'carious' and
[vaku6zo.] 'vacuousness'. The final vowel of forms (19c-d) will appear in
derived forms because they belong to the noun stem, i.e. (kari + ad + 0.) and
(vaku + oz + 0.). Indeed, the morphological structure of forms illustrated in
(19c-d) are (kari + i) and (vako. + 0.). That is, the final vowel of the noun
stem is followed by an identical vowel which corresPonds to the gender suffix.
Forms (19c-d) are realized with a posttonic single vowel as a consequence of
the establishment of a right-headed internuclear.governing relation. More
precisely, the final nuclear position - which corresponds to the gender suffix
- governs the final nuclear position of the noun stem. A light diphthong whose
segmental sequence is a sequence of either of the simplex charmless segments
i/u is realized as a single vowel (c.f. (18». In this section we considered
forms in which sequences of high voweis occupy the final nuclear positions in
antepenultimate stressed nouns.
8.2.2. Primary Stressed High Vowels

In this section we will consider forms presenting sequences of high vowels in
which one bears primary stress. First we will consider forms where a primary
stressed high vowel is preceded by another high vowel. Then, we will discuss
forms in which a primary stressed high vowel is followed by another high·
vowel. Consider (20).
(20) a. [suisa] [swisa] 'Swiss'
b. [~uizu] [~wizo.] 'judgment'
c. [siumi] [sjumi] 'jealous'
d. [miudu] [mjudo.] 'little'
(20a-b) illustrate cases where a primary stressed high front vowel is preceded
by a high back vowel. (20c-d) illustrate cases where a primary stressed high
back vowel is preceded by a high front vowel. Forms in (20) show that a
primary stressed high vowel may either be preceded by another high vowel (cf.
11'
11 In derived forms the gender suffix follows the derivative suffix.
12 These forms may also be pronounced as [karjado.] and [vakwozo.], i.e.

wi th a glide-vowel sequence. The derivation of these forms was given in
Chapter 6 where we showed that glide-vowel sequences in BP are derived from
the establishment of a right-headed internuclear governing relation.
123
.: .
forms on the left) or by a glide (cf. forms on the right) . In (21) we

illustrate the derivation of the forms given in (20a) and (20c), Le. [suisa]
- [swisa] 'Swiss' and [miudii] - [mjudii] 'little' •
(21) a. [suisa]
/ \ / \
s w s w
0 R 0 R 0 R 0 R 0 R 0 R
1 I I I I I I I I
I I I I I I I I I
I I I
N N I N I N N I N
1 I 1 I I I I I
I I I I I I I I
X X
,
X X X ---> X X X X X
I I I I I I I I I
I I I 1 I I I 1 I I
S ii i s a s u i s a
nO 1°1 !f
I
1°
I
I I
...+ ...+ ...l
b. [miudii]
/ \ / \
s w s w
0 R 0 R 0 R 0 R 0 R 0 R
I I I I I I I I
I I I I I I I I
I I
I N N N I N N N
I 1 I I I
I I I I I
X X X X x --> x x x x, X
I I I I I I I I
I I 1 I I I I I I
m i u d ii m i u d ii
1° !f, 1°
I
!f
I
I I
...1+ ...+ ...+
c. [swisa]
/ \ / \
! 5 W 5 W
0 R 0 R 0 R 0 R 0 R
,
1
I
I
1
1
I
I
I
I
I
I
1
I
I
I
·1
1
I I I I
N N N N N
,
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
1
I
I
I
I
X
I
I
X
I
I
....,
X
I
, X,
X
I I
-->x
I
I
x
/ \ 1
x
I
X
I
I
5 ii i 5 a s ii i s a
nO 1° nO 1°
I I
I I
...+ ...+
124
d. (mjlidii]
/ \ / \
! s w s w
0 R 0 R 0 R 0 R 0 R
I I I I I I I I 1
I I I I 1 I I I I
I I I I
I N N I N I N I N
I I I I I· I I I I
I I I 1 I I I 1 I
X X X X X -->X X X X
I I I I I I
I I I I 1 I / \ II II
rn i u d· -0. -m . . i-u . d· ...0.
IO !f I O !f
I I
I
1+
J.+ J.
Forms presenting sequences of vowels and forms presenting glide-vowel

sequences are derived from a sequence of strictly adjacent nuclear positions.
Therefore, the lexical representation of forms such as [suisa] - [swisa]
'Swiss' and [milido.] - [mjlidii] 'little' present a sequence of strictly adjacent
nuclear positions. This is illustrated in the. representations on the left in
(21). The derivations in (21a-b) illustrate cases in which a sequence of
vowels occurs, i.e. [suisa] and' [milido.]. We propose that pretonic nuclear
heads filled with the simplex charmless segments i and y receive the ATR
element yielding the phonetic manifestation of the tense high vowels [i] and
[u]. Our proposal follows from the fact that pretonicnuclear fieads must be
filled with positively charmed segments (cf. Chapter 3). Given that positively
charmed segments cannot be governed (cf. Chapter 1), no governing relation
holds between the adjacent nuclear posi tions in (2la-b). Therefore, a sequence
of adjacent nuclear positions is allowed to occur. The derivations in (21c-d)
illustrate cases in which a glide-vowel sequence occurs, i.e. [swisa] and
[mjlidii]. We propose that glide-vowel sequences in BP are derived from the
establishment of a right-headed internuclear governing relation yielding a
light diphthong (cf. Chapter 6). Let us now consider cases in which a Pnimary
stressed high vowel is followed by another high vowel. Consider (22).
(22) a. [hiw] 'river'
b. [periwdo.] 'period'
c. [fllij dii] 'fluid'
d. [gratlijto.] 'free'
Forms in (22) illustrate vowel-glide sequences. We propose that vowel-glide
sequences in BP are derived from the establishment of a left-headed
internuclear governing relation yielding a h~ diphthong (cf. Chapter 7).
In (23) we illustrate the derivations of the forms [hiw] 'river' and [fllijdfi]
'fluid' - which present vowel-glide sequences.
13 Forms in (22) may present an alternative pronunciation where the glide

is phonetically manifested as a high vowel. These alternative pronunciations
will be presented and discussed in the following pages.
125
(23) a.
I
I !
S W
0 R 0 R 0 R
I I I I I
I I I I I
I
I N N I
I N
I I I I
I I I I \
X X X -->x x X
I I I I I
I I I I I
h i u h i u
rI
~ r
I
~
I
1+
~ +
b.
I
I ! / \
S W S W
0 R 0 R 0 R 0 R 0 R
i\ I
I
I
I
I
I :\ i\ I
I
I
I
I
I \: N N N I
I \ N\ I
I N
I I I I I
I \ I I I \ \ I
X X X X X x ---) x x x x x X
I I I I I I I I I I
I I I I I I I I f r
f 1 u i d u f 1 u i d u
~ 1° ~ 1°
f I
I I
*+ ~
The representations on the left in (23) illustrate the establishment of a

left-headed internuclear governing relation yielding a heavy diphthong - as
illustrated in the representations on the right in (23). Forms which present
posttonic postvoca1ic glides preceded by a high vowel - as illustrated in (22)
- may present an alternative pronunciation in which the glide is phonetically
manifested as a vowel. This is illustrated in (24).
(24) a. [hiw] [hili] 'river'
b. [periwdu] [periudu] 'period'
c. [flujdu] [f1uidu] 'fluid'

d. [gratujtu] [gratuitu] 'free'
The data in (24a-b) show that forms presenting a primary stressed high front
vowel followed ~ a back glide, may have an alternative pronunciation where
the glide is manifested as a lax high back vowel. The data in (24c-d) show
that forms presenting a primary stressed high back vowel followed by a palatal
glide, may have an alternative pronunciation in which a sequence of vowels -
rather than a vowel-glide sequence - occurs. Notice that in the forms on the
right in (24c-d), primary stress does not fallon the high back vowel - as
illustrated in the forms on the left - but rather the high back vowel occurs
126
in pretonic position. 14 Let us consider the forms of (24) in more detail.

Vowel-glide sequences in BP are derived from a sequence of adjacent nuclear
positions. Therefore, forms on the left in (24) - which present vowel-glide
sequences -are derived from a sequence of adjacent nuclear positions. The
derivation of these forms was given in. (23). Let us now consider the
derivation of forms on the right in (24), i.e. where the glide is manifested
as its corresponding high vowel. First we will discuss forms illustrated in
(24a-b), Le. where a posttonic postvocalic back glide is phonetically
manifested as a lax high back vowel, e.g. [hiii] 'river'. We assume that in the
----proeess-e-f-deri-v-iftg-t-he-fe-Fm-[-h-i-u.-}---wh-i-eh-p-r-esen-ts-a-sequence of hi g~owe-1s---
- a governing relation between adjacent nuclear positions is prevented from
holding. ThiS process is illustrated in (25).
In (25) we have a sequence of adjacent nuclear positions where the far left
nuclear position (which is filled with the positively charmed segment !) has
. the property of governing the immediately following nuclear position (which
is filled with the simplex charmless segment y). Therefore, a governing
relation - more precisely a left-headed internuclear governing relation -
should hold between the adjacent nuclear positions in (25). This is not the
case.
The reasons why a governing relation does flot hold between the adjacent
nuclear positions in (25) are still obscure. What is important to mention
. .
is that such a property, i.e. a simplex charmless segment not governed by a
posi ti vely charmed segment, is restricted to sequences of adjacent. high
vowels. Forms presenting posttonic postvocalic glides where the glide is
14 Forms in (24c-d) may also be pro~nced as [flwidii] and [gratwitu],

i.e. where a glide-vowel sequence is phonetically manifested. The derivation
of these forms was given in Chapter 6 where we showed that glide-vowel
sequences are derived from the establishment of a right-headed internuclear
governing relation.
15 Unfortunately, for the time being we can only observe this phenomenon
without providing an explanation for it.
127
preceded by a non-high vowel, e.g. [paw] 'stick'; [dewza] 'goddess', do not

present an alternative pronunciation where the glide is phonetically
manifested as a lax high vowel, i.e. *[pau] and *[deuza] are not possible
forms in BP.
Let us now discuss forms illustrated in (24c-d), e.g. [flujdu] - [fluidu]
I fluid I . In these forms a primary stressed high back vowel followed by a
palatal glide presents an alternative pronunciation in which a sequence,o!
vowels - rather than a vowel-glide sequence - is phonetically manifested.
Forms presenting a vowel-glide sequence, e.g. [flujduJ, are derived by the
establishment of a left-headed lnternuclear governing relation so that a
vowel-glide sequence is realized. (cf. (23b)). Forms presenting a sequence of
vowels, e.g. [fluidu], illustrate the case where a pretonic nuclear head
filled with the simplex charmless segment y acquires the ATR element. It is
thus manifested as a tenne high vowel, i.e. [u]. A sequence of vowels, i.e.
[fluidul, is manifested.
We expect that forms presenting a primary stressed high back vowel followed
by a palatal glide, e.g. [flujdu], will present an alternative pronunciation
in which a sequence of vowels is phonetically manifested, i.e. [fluidu]. This
is exactly the case. It remains to be seen why forms presenting a high back
vowel followed by a primary stressed high front vowel, e.g. [~uizul
'judgment', cannot be phonetically manifested as' a vowel-glide sequence, i.e.
* [~ujzu] .
8.3. Conclusion
In this Chapter we considered forms presenting intervocalic glides. We showed
that intervocalic palatal glides are derived from a nuclear position filled
with the simplex charmless segment [i]. We also considered forms presenting
a sequence of high vowels, and showed that whether a high vowel or a glide is
phonetically manifested depends on the governing relations established between
adjacent nuclear positions.
We have seen that the simplex charmless segment 1 may govern the simplex
charmless segment y and vice-versa (cf. [sabju] and [tenwi] in section 8.2.2).
Considering the definition of government, i.e. a binary, asymmetric relation
between adjacent positions~. we do not. expect a governing segment to be
governed. This is because if a given segment may either govern or be governed,
the aSYmmetric nature of government is violated. Further research is still
needed for a better understanding of the governing properties of simplex
charmless segments and the asymmetric nature of government. The process of
vowel-glide alternations presented in this Chapter aims to throw some light
on the distrussion of this issue.
16 A form such as [fluidu] may also be pronounced as [flwidu]. Forms

presenting a glide-vowel sequence are derived from the establishment of a
right-headed internuclear governing relation yielding a light diphthong (cf.
Chapter 6).
128
CONCLUSION
CONCLUSION
In this thesis we have considered phonological phenomena involving nuclear
positions. Specifically, we considered vowel coalescence in BP with emphasis
on the processes of vowel-glide alternations. .
We defined how word stress is assigned in BP showing that primary stress is

lexically determined where feet may be either binary or ternary. We have also
claimed that metrical trees are built onto the level of nuclear projection and
we stated the metrical constraints imposed on posttonic constituents in BP.
We have shown that the occurrence of certain vowels in BP depends on their
position with respect to primary stress. We thus proposed that the
distribution of vowels in BP is accounted for by charm constraints imposed on
nuclear segments.
A review of the literature on the process of vowel-glide alternations in BP
was presented showing that earlier works encounter problems in explaining the
conditions under which a high vowel alternates or fails to alternate with its
corresponding glide, and in determining the phonological representation of
diphthongs in BP. Therefore, further research on the process of vowel-glide
alternations in BP is still needed.
Our proposal which accounts for the process of vowel-glide alternations in BP
was then presented. Such a proposal is an expansion of the earlier treatments
of governing relations between adjacent nuclear positiuns. We ciaimedthat in
a sequence of adjacent nuclear positions the Obligatory Contour Principle
(OCP) applies under government.
Forms presenting prevocalic glides, Le. [jV,wV], and forms presenting
prevocalic high vowels, i.e. [iV,uV], were discussed. Evidence was provided
showing that glide-vowel sequences in BP correspond to the representation of
a light diphthong. Light diphthongs in BP are derived from the establishment
of a right-headed internuclear governing relation. Prevocalic high vowels
represent the cases where an internuclear governing relation is prevented from
holding.
We also considered forms in which a velar stop-glide sequence occurs, i.e.
~. The back glide in BP may cor~espond either to the element uO
syllabified in the representation of a light diphthong or to the element uO
together with a velar stop consonant syllabified in the representation of a
complex consonant. We claimed that when the back glide is phonologically
interpreted as the element uO syllabified in the representation of a complex
consonant, i t corresponds to the cases where the element uO is lexically
associated to a position other than~ nuclear head.
Forms presenting postvocalic glides, Le. [Vj,Vw], and forms presenting
postvocalic high vowels, i.e. [Vi,Vu], were then considered. Evidence was
provided showing that vowel-glide sequences in BP correspond to the
representation of a heavy diphthong. Heavy diphthongs in BP are derived from
the establishment of a left-headed internuclear governing relation.
Postvocalic high vowels represent the cases where an internuclear governing
129
.: .
CONCLUSION
relation is prevented from holding.

Finally, we analysed forms presenting intervocalic glides and forms presenting
sequences of high vowels. Intervocalic palatal glides are derived from a
nuclear position filled with the simplex charmless segment 1. Concerning forms
which present sequences of high vowels we determined which of the high vowels
in the sequence may be realized as a glide. The manifestation of a high vowel
or a glide depends on the governing relations that adjacent nuclear positions
contract with each other.
Our research provides a general account of the occurrence of high vowels and
glides and determines how glides are phonologically interpreted in BP.
130
REFERENCES
Almeida, A. (1976). "The Portuguese Nasal Vowels: Phonetics and Phonemics".

Readings in Portuguese Linguistics. ed. by Jurgen Schnidt-Radefe1t.
North-Holland Linguistic Series 22. 349-396
Alvarenga, D. et a1 (1989). Sobre a Reducao do Ditongo. Projeto de
A1fabetizacao. Faculdade de Letras. UFMG •. Belo Horizonte.
Beddor, P.S. (1982). Phonological and Phonetic Efects of "Nasalization on V~wel
Height. Ph.D. University of Minnesota.
Bisal, L. (1989). "0 Ditongo na Perspectiva da Fonologia Atual". in
D.E.L.T.A., Volume 5, Number 2. 185-224.
Booij, G. (1989). "On the Representation of Diphthongs in Frisian". Journal
of Linguistics. 25. 319-332.
Cagliari, L.C. (1977). An Experimental Study of Nasality with Particular
Reference to Brazilian Portuguese. Ph.D. dissertation. Edinburgh
University.
Cagliari, L.C. (1982). Elementos de Fonetica do Portugues Brasileiro. IEL.
UNICAMP.Sao Paulo.
Camara (Jr), J.M. (1953). Para 0 Estudo da Fonemica Portuguesa. Organizacoes
Simoes. Rio de Janeiro.
Camara (Jr), J. M. (1970). Estrutura da Lingua Portuguesa. Vozes. Petr6polis.
Rio de Janeiro. 14th edition (1984).
Carreira, M. (1988). "The Representation of Diphthongs in Spanish". Studies
in the Linguistic Sciences. Volume 18. Number 1. 1-24.
Cavaco, A. M., (in preparation). Os Padr5es das Alternancias Vocalicas e da
Vogal Zero na Fonologia Portuguesa. Ph.D. dissertation. Universidade
dos Acores.
Charette, M. (1988). Some Constraints on Governing Relations in Phonology.
Ph.D. dissertation, McGill University.
Charette, M. (1991). Conditions on Phonological Government. Cambridge
University Press.
Goldsmith, J. (1979). Autosegmental Phonology. PhD Dissertation. Distributed
by IULC. New York. Garland Press .
..,.
Goldsmith, J. (1990). Autosegmental and Metrical Phonology. Basil Blackwell.
Halle, M. (1986). On the Phonology-Morphology Interface. Ms. MIT.
Harris, J. (1983). "Syllable Structure and Stress in Spanish: a Nonlinear
Analysis. Linguistic Inguiry Monograph. 8. MIT Press.
131
Harris, J. & Kaye, J.D. (1988a). "A Tale of Two Cities: London Glottalling and
NYC Tapping". London Phonology Seminar. SOAS_
Harris, J. & J. KAYE. (1988b). "A Government-Based Approach to Lenition".
London Phonology Seminar. SOAS.
Harris, J. (1990). "Segmental Complexity and Phonological Government".
Phonology Yearbook 7.2. 255-300.
Head,B.F. (1964). A Comparison of the Segmental Phonology of Lisbon and Rio
de Janeiro. Ph.D. dissertation. Austin. Texas.
Istre, G.L. (1971). A Phonological Analysis of a Portuguese Dialect. Louisiana
State University and Agricultural and Mechanical College. Ph.D.
dissertation.
Jeroslow, E.H.M. (1974). Rural Cearense Portuguese: A Study of the Variety of
Non-standard Brazilian Speech. Ph.D. dissertation. Cornell University.
Kaye, J.D. (1982)."Harmony Processes in Vata". in The Structure of
Phonological Representations (Part II). H. van der Hulst & N. Smith
(eds). Foris Publications.
Kaye, J.D. (1985). "On the Syllable Structure of Certain West African
Languages". in D. Goyvaerts (ed.), African Linguistics: Essays in
Memory of M.W.K. Semikenke, 285-308. Amsterdam, J. Benjamins.
Kaye, J.D. (1989a). "Raddopiamento Sjntattico Norwegian Style". Paper
presented at the London Phonology Seminar. SOAS. London.
Kaye, J.D. (1989b) .. Phonology: a Cognitive.View Lawrence Erlbaum Associates,
Publishers. N.J.
Kaye, J.D. (1989c). ""Coda" Licensing".Phonology Yearbook:7.2. 301-330.
Kaye, J.D. (1989d). "On the Interaction of Theories of Lexic~l Phonology and
Theories of Phonological Phenomena". Phonologica 1988.
Kaye, J.D. (198ge). "Current Issues in Phonology". SOAS Lectures. London.
Kaye, J.D. (1990a). "Government in Phonology: The Case of Moroccan Arabic". The
Linguistic Review, 6: 131-160. "1986/1987" (1990).
Kaye, J.D. (ed). (1990b). Phonology Yearbook 7.2.
Kaye, J.D. & J. Lowenstamm (1984). "De la Syllabicite". in Forme sonore du
langage. F. Dell, D. Hirst et J .. -R. Vergnaud (eds.). Paris, Hermann.
Kaye, J.D., Lowenstamm, J. & J-R. Vergnaud (1985). "The Internal Structure of
Phonological Elements: A Theory of Charm and Government". Phonology
Yearbook 2 (1990). 305-328.
Kaye, J.D., Lowenstamm, J. & J.-R. Vergnaud. (1990). "Constituent Structure
and Government in Phonology". Phonology Yearbook. 7.2. 193-231.
132 . -
..... '
".-.
- :... ...:.:~~
Kaye, J.D. & J-R Vergnaud (1990). "On the Interaction of Theories of Lexical
Phonology and Theories of Phonological Phenomena" Paper presented at
GLOW -Cambridge. To appear in Phonologica 1988. Cambridge University
Press.
Kean, M.-L. (1974). "The Strict Cycle in Phonology". Linguistic Inguiry 5.
Kenstowicz, K. & J. Rubach. (1987). "The Phonology of Syllabic Nuclei in
Slovak". Language, 63. 463-497.
Leben, W. (1973). Suprasegmenta1 Phonology. Ph.D. dissertation. MIT.
,.
Lopez, B.S. (1979). The Sound Pattern of Brazilian Portuguese (Cariocan .......,~
Dialect). Ph.D. dissertation. University of California. Los Angeles.
Lowenstamm, J. (1986). "A Propos dtune Hypothese sur la Forme Primitive du
Type B en Amharique". in Revue Quebecoise de Linguistigue. 16. 157-180.
Madureira, E.D. (1987). Sobre as Condicoes da Voca1izacao da Lateral Palatal
no Portugues. Master Thesis. UFMG. Belo Horizonte.,-_
Maia, E.A. (1981). Hierarquias de Constituintes em Fonologia. Anais do V
Encontro Nacional de Linguistica. PUC. RJ.
'.
Major, R. (1985). Stress and Rhythm in Brazilian Portuguese. LanqUage~ 61,
Number 2.
Major, R (1981). Stress-timing in BP. Journal of Phonetics. Number 9.
Marota, G. (1978). The Italian Diphthongs and the Autosegmenta1 Framework. In
Bertinetto,P.M. and Loporcaro, M. (ed). Certamen Phonologicum. Papers
from the 1987 Cortona Phonological Meeting. Rosenberg and Sell ier.
Turin.
McCarthy, J. (1986). "OCP Effects: Gemination and Antigemination". Linguistic
Inquiry. Number 17.
Naro, A.J. (1971). "Resolution of the Vocalic Hiatus in Portuguese: Diachronic
Evidence .for.. .Binary Feat.ures" .• Language. Volume.A 7. Number 2..
Nunes, J.J. (1951). Compendio de Gramatica Hist6rica. 4. Editora Lisboa.
L~vraria Classica Editora.
Odden, D. (1988). "On the Obligatory Contour Principle". Language. Number 62.
Oliveira, M.A. (1983). "Phonological Variation and Change in Brazilian
Portuguese: The Case of the Liquids". Ph.D. Thesis. University of
Pennsylvania.
Redenbarger, W. J. (1976). Portuguese Vowel Height and the Phonological
Theory: A Generative Re-analysis Based on Tongue-root Features. Ph.D.
Harvard University.
Segundo, s.o. (1990) "On Empty Nuclei and Metrical Structure in Brazilian
Portuguese". SOAS Working Papers in Linguistics and Phonetics. Number
1. University of London.
133
·: .
Segundo, S.O. (in preparation). Stress and Related Phenomena in Brazilian

. (Natal) Portuguese. Ph.D. Thesis. SOAS.",
Shaw, I.S. (1986). Vowel Nasality in Brazilian Portuguese: An Experimental
Approach with Focus on Derivational and Inflectional Alternations.
Ph.D. dissertation. University of Kansas.
Silva, T.C.A. da (1989a). "Vowel Nasalization in Brazilian Portuguese". ms.
SOAS.
Silva, T.C.A. da (1989b). "Branching Onsets in Brazilian Portuguese". ms.
SOAS.
Silva, T.C.A. da (1989c). "On Empty Nuclei in Brazilian Portuguese". ms. SOAS.
Silva, T.C.A. da (1992). "The Phonological Representation of Velar Stop-Glide
Sequences". SOAS Working Papers in Linguistics and Phonetics. Number 2.
University of London.
Steriade, D. (1984). "Glides and Vowels in Romanian". BLS Proceedings. 10.
Berkeley University Press. 47-54.
Vandressen, P. (1975). "0 Vocalismo Portugues: Implica~oes Te6ricas. Revista
Brasileira de Linguistica. 2. 80-103.
Vergnaud, J .-R. (1982). "On the Foundations of Phonology". Paper presented at
the 1982 GLOW Colloquium, Paris.
Yip, M. (1988). "The Obligatory Contour Principle and Phonological Rules: A
Loss of Identity". Linguistic Inquiry. Number 19.
Yoshida, S. (1990). A Government-Based Analysis of the 'Mora' in Japanese.
Phonology Yearbook 7.2. 331-335.
Yoshida, S. (1991). Some Aspects on Governing Relations in Japanese Phonology.
Ph.D. dissertation. SOAS. London.
134
View publication stats

1992 TESE NuclearPhenomenainBrazilianPortuguese

Uploaded by

Copyright:

Available Formats

You might also like

1992 TESE NuclearPhenomenainBrazilianPortuguese

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1992 TESE NuclearPhenomenainBrazilianPortuguese

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Nuclear Phenomena in Brazilian Portuguese

Thesis · January 1992

Thais Cristófaro Silva

The user has requested enhancement of the downloaded file.

THAIS CRISTOFARO ALVES DA SILVA

A THESIS SUBMITTED TO THE DEPARTMENT OF

SCHOO~ OF ORIENTAL AND AFRICAN STUDIES

NUCLEAR PHENOMENA IN BRAZILIAN PORTUGUESE

In this thesis we will analyse phonological phenomena which involve nuclear

Department of Phonetics and Linguistics

~ Thals Crist6faro Alves da Silva

c/o Departamento de Linguistica

1.0. Introduction ------------------~---------------------------------- 1

----l~veriiIng Properties of Nuclear Segments ----------------------~--

CHAPTER 2: WORD STRESS IN BRAZILIAN PORTUGUESE

2.0. Introduction ----------------------------------------------------- . 17

CHAPTER 3: THE VOWEL SYSTEM OF BRAZILIAN PORTUGUESE

CHAPTER 4: REVIEW OF THE LITERATURE

6.0. Introduction ----------------------------------------------------- 67

CHAPTER 7: POSTVOCALIC GLIDES AND POSTVOCALIC HIGH VOWELS

7.0. Introduction ------~----------------------------------------------,98

CHAPTER 8: INTERVOCALIC GLIDES AND SEQUENCES OF HIGH VOWELS

8.0. Introduction ----------------------------------------------------- 109

1 The translation into English of the data originally collected in

In Chapter 5 we consider governing relations involving a sequence of strictly

THE THEORY OF CHARM AND GOVERNMENT

1.1. A Theory of Phonological Strings

1 For other lines of research in GP and a full presentation of the theory

the nature of headship within a governing domain. 2

2 Directionality of goverrunent in phonological constituents differs from

Nuclear posiVons and their segments are lexically associated to nuclear

(2) illustrates a sequence of skeletal positions which are associated to

6 This property follows from the "Licensing Principle~ which establishes

9 This claims follows from the "Binary Theorem" which is demonstrated in

presents branching onsets, ego [klaru] 'clear' or [livru] 'book' .10

10 A phonological process involving branching onsets in BP~s discussed

A governing relation involving two adjacent positions which are associated to

(9) illustrates a non-nuclear interconsti tuent governing domain. According to

13 A governing relation between an onset and a rimal complement still

In order to consider the representation illustrated in (10) - which involves

Let us now examine the governing relations which involve a sequence of

1.2.The Internal Representation of Nuclear Segments

17 In Chapter 5 we provide a theoretical discussion on governing

20 In Chapter 6 we will address in detail the representations in which

(19) illust!rtes a phonological expression whose head is another phonological

23 The positively charmed element ..+ anar It have the property of

25 For descriptions and analysis of nasal vowels in BP see CAMARA

representation is (VO.~f)+. Given that the segment [a] is positively charmed

1.3. Governing Properties of Nuclear Segments

a skeletal position ta govern or to be governed by an adjacent skeletal

~ 27 We will restrict the discussion of governing properties of nuclear

CHAPTER 1: THE THEORY OF CHARM AND GOVERNMENT

Simplex Neutrally Charmed Segments

CHAPTER 1: THE THEORY OF CHARM AND GOVERNMENT

In (25) we illustrate the governors and the governees of nuclear segments in

2.1. Primary Stress Assignment